Ted Hallum: Meet Your Host

Show Notes

In this episode, former guest and fellow Veterans in Data Science & Machine Learning community member, Chris Sanchez, provides a formal introduction to your host here on The Data Canteen, Ted Hallum.

FEATURED GUEST:

Name: Ted Hallum

LinkedIn: https://www.linkedin.com/in/tedhallum/

Email: founder@vetsindatascience.com

SUPPORT THE DATA CANTEEN (LIKE PBS, WE'RE LISTENER SUPPORTED!):

Donate: https://vetsindatascience.com/support-join

EPISODE LINKS:

Self-driving Cars with Ducky Town (edX MOOC): https://bit.ly/3jPWcQy

Becoming a Data Scientist (podcast): https://apple.co/3jQLIjT

Eye on AI (podcast): https://apple.co/3qpWlwu

SuperDataScience Podcast: https://apple.co/3u2JaUs

Build a Data Science Career (book): https://amzn.to/2NaGV0H

AI Superpowers (book): https://amzn.to/3s1p8aT

PODCAST INFO:

Website: https://vetsindatascience.com/thedatacanteen

Apple Podcasts: https://podcasts.apple.com/us/podcast/the-data-canteen/id1551751086

YouTube: https://www.youtube.com/channel/UCaNx9aLFRy1h9P22hd8ZPyw

Stitcher: https://www.stitcher.com/show/the-data-canteen

CONTACT THE DATA CANTEEN:

Voicemail: https://www.speakpipe.com/datacanteen

JOIN THE VDSML COMMUNITY:

LinkedIn: https://www.linkedin.com/groups/8989903/

OUTLINE:

00:00:07​ - Introduction

00:01:05 - A Not-So-STEM Background

00:09:32​ - Military Experience

00:11:33 - Post-military, Pre-data Science Civilian Career

00:14:06​ - How I Bumped Into Data Science

00:16:28​ - Time to go to Grad School

00:22:12 - Greatest Motivation

00:23:43 - Grappling with Imposter Syndrome

00:28:08 - Favorite Courses in Grad School

00:30:24 - The Antidotes to Imposter Syndrome

00:31:26 - First Data Science Job

00:42:42 - Second Data Science Job...err uhh...Geospatial Python Developer

00:48:58 - Machine Learning Engineer

00:51:19​ - Veterans in Data Science & Machine Learning and The Data Canteen

00:57:08 - Favorite Programming Language

01:00:41​ - Next MOOC: Self-Driving Cars with Duckietown

01:04:46​ - Podcast Recommendation #1: Becoming a Data Scientist

01:05:24 - Podcast Recommendation #2: Eye on AI

01:06:16 - Podcast Recommendation #2: SuperDataScience Podcast

01:06:41​ - Book Recommendation #1: Build a Data Science Career

01:07:15​ - Book Recommendation #1: AI Superpowers: China, Silicon Valley, and the New World Order

Transcript

DISCLAIMER: This is a direct, machine-generated transcript of the podcast audio and may not be grammatically correct.

Ted Hallum: [00:00:07] Welcome to episode four of The Data Canteen, a podcast that focuses on the care & feeding of data scientists and machine learning engineers who share in the common bond of U.S. Military service. Today, I'm joined by fellow community member Chris Sanchez, who's kindly agreed to introduce me to you. Hope you enjoyed the conversation as much as I did.

Hey Chris, welcome back to the show. I appreciate you coming on and being willing to interview me and introduce me to the Veterans in Data Science and Machine Learning Community.

Chris Sanchez: [00:00:35] Ted. I've been looking forward to this for quite some time. So, without further ado, let's jump right into it. So where are you from originally?

Ted Hallum: [00:00:44] I'm  originally from Georgia.

Chris Sanchez: [00:00:45] Okay. So the only experience I have with Georgia is spending three weeks in Fort Benning. Going through jump school, which I remember being a very miserable experience. I hope yours wasn't the same when you were growing up there. So, given you what you're doing now were you particularly interested in STEM at an early age?

Did that form any part of your studies in college?

Ted Hallum: [00:01:05] I'm glad you asked me that because this is a message that I want to make sure I get across to anyone who's considering data science and machine learning, or maybe you've heard of it and you think that would be cool, but your background is not heavily STEM or heavily quantitative.

No. In high school I struggled in math. I went to a rural school that didn't have the best teachers. My folks didn't have heavy quantitative backgrounds either. So it was just this perfect storm to not get involved in math.

I did the minimum of math in high school to go to the college that I knew I wanted to go to.

And then I did a degree program that I knew wouldn't require any math. Then, went into a career that for the most part didn't need math. I think I remember you saying something similar about your military career. I didn't need a lot of math for that. But then when I found out about data science and I did need math. So, I had to come face to face with this fear that I'd had all the way back from when I was a teenager.

That's when I discovered, and you and I talked about this at length when you were on the show last time,  it's a whole different world now. It's not like it was when we were in high school, back in the early two thousands, late nineties, whatever. There's Khan Academy, there's YouTube, there's communities where you can get involved with learning statistics, linear, algebra, whatever it is that you need to know with other people who can help you. It can be a collaborative learning experience.

So no, I did not have a strong STEM or quantitative background, and it was a lot of work. I don't want to gloss over that fact. Going back and picking up these skills as an adult was not by any means easy, but it's certainly doable. So, if this is something you really want and you're willing to put in the work, I don't think it's anything you need to worry about. You certainly shouldn't let it hold you back.

Chris Sanchez: [00:02:49] So you said something earlier, you said you struggled with math. Was it something that, you conceptually struggled with? Like it was hard for you to grasp the concepts or was it something that, you know, because of the,  the environment that you were raised in that it was something that you just weren't even interested in to begin with?

Ted Hallum: [00:03:05] I would say probably at least a little bit of both. I think it was more the latter. I think it was the way that the concepts were introduced to me. It wasn't a state-of-the-art introduction in the first place. Then, what made it worse is that maybe I wasn't the most mathematically inclined person . But, fast forward all these years and get on Khan Academy and I listen to Sal Khan, who is just, he is an extremely gifted mathematics teacher.

Chris Sanchez: [00:03:28] Yeah.

Ted Hallum: [00:03:29] And, it was a totally different experience. Picking up those concepts was so much easier when you could listen to them be explained by somebody who can explain it artfully.  Once I got access to that kind of resource, then I just sprouted wings and took off.

Chris Sanchez: [00:03:47] If someone had told your high school self that you would be, in a career that involves heavy amounts of linear algebra, multi-variable calculus, what would you have said to them?

Ted Hallum: [00:03:56] That would not be possible, and I could never do that.

 You've probably heard people say to approach a big problem, like you would an elephant and then eat it one bite at a time, and that's absolutely how I had to do it. I just had to put on blinders and just reached for one bite and eat that one bite at a time.

Because if I had looked at the full scope of what was in front of me and everything that I needed to learn, it would've just been overwhelming, and I probably would've given up. The right approach was to just say, okay, today I've got this in front of me. This is the skill, or this is the concept, that I need to master.

I'm going to focus on this and do that today, and then let that naturally lead to the next concept that you need to learn tomorrow.  Then, pretty soon, 90 days has gone by, or 120 days has gone by and you look back and you're like, wow. Like, I blasted through an entire advanced statistics and probability course.

Now, I'm good on that. And you move into the next thing. So , I think that's the best way to go about it. 

Chris Sanchez: [00:04:55] Something you said made me think back to my time, learning calculus in college. I remember at the time, not being able to put it into place of  where am I going to use this in the future?

So it wasn't that interesting to me. There was a challenge and taking it and then okay where am I going to apply it? And that mindset was completely different when I learned for data science, because I was like, "Oh, I need to know why to limit works because that's going to be a foundation for derivatives, and that's important because that's the foundation for backpropagation." So, being able to take these sometimes esoteric concepts into context, I found is extremely useful in that motivation behind wanting to learn it in the first place.

Ted Hallum: [00:05:37] I think another thing that's worth pointing out too, because the Veterans in Data Science Machine Learning community is it's a group of people that run the full spectrum. You have everything from people in maybe analyst roles that are aspiring to just towards data science in general.

Then, within the data science portion of our audience, you have people who only want to do the applied side all the way up to people who have PhDs that are very much into the theoretical side and developing new algorithms So, there's a big gamut. I think probably the sweet spot that maybe encompasses the biggest portion of our listeners would be people who are doing applied data science .

For those folks, or if you're on the track towards that, if you're in a master's degree program headed towards applied data science, you definitely need to, as a foundation, know these mathematical concepts, they are going to inform which algorithm you choose. And then ,they're going to also inform the way that you use that algorithm.

Just as an example from this past week, someone that I was talking to about at the concept of Principle Component Analysis, also known as PCA. I hadn't used a PCA model in a while and I thought I should go brush up on that and just refresh my mind on PCA. So I pulled up a quick article and I was reading through it.

And it pointed out that if you're going to use a PCA model, that it's important because the mathematics behind the model that you make sure that you normalize or standardize your data. Because if you have one variable in your data that, only ranges between zero and one, like a percentage and then you have another variable that's like age.

So let's say that  the one variable is how far along you are in grad school. And it's somewhere between zero and 0.999, whatever. And then there's the age column. Most people are somewhere between 18 and 80, probably if you're in our audience . The way that PCA works, it would always put all the importance on the age column because the values and the age column are so much larger than the values in that percentage column.

So you would need to know that you need to know that mathematical background, because that tells you how you need to pretreat your data before you actually feed your data into the PCA model in order for the PCA model to perform correctly. And, yeah, you don't actually have to do the math because there's a Python package.

You're probably using scikit-learn or whatever it's going to implement the math, but you have to understand enough about how the algorithm works to do some fundamental things, to get the result that you're actually after.

That's the one thing I'd point out to people who are headed towards applied data science, is that, yes, you're going to have to do this math. Yes, you're going to have the background and the background will continue to help you as you leverage these statistical and quantitative approaches. If nothing else to help you prepare your data properly. So you get the right output. But you're probably not going to have to know the in-depth guts of, for instance, how PCA works to use PCA.

So just know that you're going to have to be, up to your elbows in the math in grad school and when you're getting started. Maybe not as much once you're out using it. But that background is going to continue to pay dividends for as long as you're in the job. On the other hand, if you're on a track towards doing theoretical work, developing new deep learning model algorithms, then that's going to be a different story.

You'd essentially be building the next evolution of scikit-learn and you definitely have to understand all the math so that you can build a Python package that other people will ultimately use.

Chris Sanchez: [00:09:02] Ted for a kid that grew up without a strong math background and is not able to break down PCA.

I think that's pretty impressive. You had an entire career in the military and now here you are, landing in this data science field. So given your background and given your, your former career walk us through that evolution. Like how did you go from, Hey, I'm the kid who struggles in math in high school to now on this intelligence analyst. And now I'm a machine learning engineer. Please explain that evolution, if you would.

Ted Hallum: [00:09:32] So it was definitely a windy journey. I'll take you all the way back to 2001. I was. In my dormitory freshman year of college, when 9-11 happened, I was about to head out for classes that day. I was going down the hallway and I remember passing by another guy.

He had a bicycle and he said, somebody just flew a plane into the world trade centers. And I remember I asking him like, "Was it some kind of crazy accident?" And he said, "No, there was two planes that crashed, it seems like it was deliberate. Like it was an attack." And  I almost went ahead and enlisted in the Army right then.

But I had a few people in my life that offered some really sage advice, and I'm so glad that I listened to them. They said, "This is going to spark a prolonged conflict. This is going to go on for a long time. If you want to contribute to the response to this attack, there's going to be plenty of time and opportunity for you to do that. But, there's so many people who leave college for whatever reason, and they're never able to get back. So, you should probably go ahead and finish college, finish your degree, and then, at that point, if that's still something that you aspire to contribute to, then you should do it." So I was like, okay, that makes sense.

I knew that one to do intelligence analysis. I knew I wanted to pursue getting a government security clearance. As soon as I graduated from college, I shipped off, did basic training, then went into my advanced individual training - AIT - for all-source intelligence analyst.

When I finished training, this was right around February of 2007. I got assigned to the Third Infantry Division at Fort Stewart, Georgia, and then I headed out on my first tour to Iraq was a 15 month tour in Baghdad.

Chris Sanchez: [00:11:07] Long tour.

Ted Hallum: [00:11:09] Yeah. I came home. We were back in the States there at Fort Stewart for about 13 months. Then, I headed out again for a second extended tour, a 13 month trip. That time I was in Diyala province on the Eastern border of Iraq and Iran.

At that time , they were still stop-lossing in people. So, I went beyond the end of my contract, and the Army ended up actually getting about four years and eight months out of me. Then, I finally came home.

Luckily on that second deployment , I had met a Civilian analyst who worked with a program called the Counter-insurgency Targeting Program, and it was based out of the National Ground Intelligence Center .

Chris Sanchez: [00:11:44] NGIC!

Ted Hallum: [00:11:45] That's right NGIC. I loved the work that he did, and I had already decided that I really wanted to work on that program. I had gotten to known him and he became familiar with the quality of work I did as an analyst. So, he was willing to put in a referral for me.

So, I got back, out processed virtually immediately, and then headed off to be Contract Intel analyst on the Counter-insurgency Targeting Program.

That actually was a really neat transition for me because my day to day battle rhythm, if you will, wasn't that different. I was still doing Intel analysis work and about half the time I was still deployed overseas to Iraq or Afghanistan with military units.

I was very fortunate too because my very first trip out with the Counter-insurgency Targeting Program, I was able to get In a position where I supported a couple of SOF elements, and that went well. So then every assignment after that, I got embedded with other SOF elements.

After that, I had a few more other cool opportunities.

A close cousin of the Counter-insurgency Targeting Program was another NGIC program called the I2 program or Identity Intelligence program. I was able to be a team lead on that and I supported SOF elements once again in that role.

For anybody who's familiar with the government contracting world, unfortunately, it was at that point that the company I was with, the contract came up for recompete, and they lost the contract. So, I had the proverbial decision of, do I sign on with the new company that just won the contract?

Or, do I stay with the company that I've had a good experience with and see what, if anything, they might have on tap to offer me. I wasn't enthusiastic about the new company that had won the contract. They didn't have the best reputation in the government contracting world. So, I decided I'm going to try to stay with the company that I had worked for as a team lead there on the I2 program.

I spoke with them and said, "Hey, I'd love to stay with you guys." They said, "Yeah, we'd love to have you stay with us too. Let's look at your background. Let's look at our current portfolio contracts and see if we can find a good match." So, it was a couple of days, they came back to me and they said, "Hey, you provided OSINT support to army cyber command. So we've got another contract here at the NGIC where we're providing Cyber All-source Analysis support. You'd be a great fit for that. Why don't you come on board and do that?" And I said, "Sure, absolutely. "

It was in that role that I started to bump up against this thing called data science and machine learning for the first time. So this is early 2018. I had never heard of it before that. The more I started to learn about it, the more I started to develop this intuition that, wow, this is a big deal.

This is going to change the way Intel analysis is done.

Chris Sanchez: [00:14:24] What did you mean by you bumped into it? People were talking about it or you were seeing feeds in your internet?

Ted Hallum: [00:14:29] I saw feeds in my internet about it on LinkedIn, but also just as I was reading stuff that pertained to cyber. At first, I just saw like the term data science.

I'm like, what's that? And I go and Google data science, and then I realized, Oh, this has been a thing since 2011. Of course I realized after that, the more I learned that the concepts behind data science have actually been around much longer than that. But the term has existed since 2011.

Then, as an extension, I learned about machine learning and deep learning and all these other things that fall underneath the umbrella of artificial intelligence. I start taking MOOCs, I did a Coursera course. I started doing Data Camp sticking my toes in the water.

Then , August of that year, so this is August of 2018. The DoDIIS conference, which is the DoD's IT tech conference that they have every year. The commanding general SOCOM at the time he gave a keynote address and the entire address was about how SOCOM had taken the 13 months before that. And they had totally retooled their Intel apparatus to make it data-driven.

For me, that was a moment of affirmation. I heard that and I was like, wow, this conviction that's been building in me for these last six or eight months. This is a real thing  because anybody that has background with the DOD or the intelligence community knows that compared to the branches of the military or the big three letter agencies, SOCOM is relatively small, and they're exceptionally well-funded.

So, whenever you see SOCOM make a hard left and a certain direction, and then put the pedal to the metal, that's like a Canary in the coal mine. That's your indicator for where all the lumbering behemoths will go when they can change their momentum and go a different direction.

Usually that takes another five to seven years in a lot of cases. So, when I heard that at DoDIIS I was like, okay I believe now that this is not just a hunch, this is a real thing. This is gonna really change the way Intel analysis is done. The way I was taught to do until analysis in five years is going to be considered the dark ages, archaic.

 So, that's when I decided. I need to go back to grad school. I want to be serious about how I get these skills. I don't have any foundation, I wasn't one of these people that already has a master's degree in engineering or a PhD in physics I need everything. I need to start with the math that I need, and I need the fundamental Python and R programming skills.

I need to go from there. So for me, the right answer was to just pursue a quantitative master's degree. That was the point when I started to apply to programs.

Chris Sanchez: [00:16:54] So you have, just to recap, you grow up in Georgia, not a great math background, you ended up pursuing a liberal arts or a humanities focused degree in college.

You choose to become an Intel analyst after 9-11, you spent almost five years active duty, and then the remainder in the declared defense contracting space. Along the way, you stumble into this data science field and you realize that is going to be the future. So, Ted decides, "Hey, that's for me.  Even though I know that I don't have an affinity for math, I'm going to go do that anyway."

To me, that's kind of like the first question is, all right, so you saw the writing on the wall, but what gave you that confidence to say, "Hey, I want to be a part of that, and I don't just want to be an analyst sitting here on the sidelines, letting things happen to me. I actually want to be a part of that revolution."

Where do you get the motivation for that?

Ted Hallum: [00:17:52] I knew I needed some help. I knew I needed some advice and mentorship. I knew that there was nobody in my physical world that had taken that path and gone from any kind of military background to data science.

I didn't know a single person actually, at that point, who was in data. So I thought, " I think what I'll do is I'll go on LinkedIn and I'll try to find some people that meet three criteria. That they have some touch point with the DOD, that they are in data science or a data science related field, and that they have a touch points with graduate education."

That was the three criteria, and I started searching. Finally, after two days, I found about four or five people. I just sent them a message and said, "Hey, my name's Ted Hallum. I have this background as a Veteran and in the cleared contracting space. I've recently learned about data science. I want to pursue it seriously. I want to go to grad school. I could use some advice and mentorship. Will you help me?" Of those five people that I messaged, three I never heard from, one person messaged me back, and they were like, "Oh yeah, I would love to help you. I'd love to pay it forward . Can you send me some more information? I'll give you some feedback." So I sent over my, I had a draft personal statement at the time for applying to grad school - never heard back from that person , which wasn't encouraging. Then,  the last person got back to me and said, "Yeah, I would love to help you!"

I told them the same thing. I said, "Well, the first thing I need help with, so I'm applying to grad school, I've got this personal statement written up. I'm going to apply to this one particular university I had in mind. And I could just use any insight you have on how I could better frame myself as a candidate for this program." So, at that point I send my personal statement over and a day goes by and I get it back and it has tremendous feedback. The person was super helpful.

I had just been looking at tons and tons of LinkedIn profiles looking for these three criteria, but probably not paying as much attention as I probably should.

One of the things that struck me is when this person got back to me and his name is Dr. Joe Wilck. When he got back to me, beneath all of the feedback he said, "I actually think you would be a fantastic candidate for my program, which is a Master of Science and Business Analytics program.  Here's the link, I recommend you go out and take a look." Immediately, there was two things. First. I was like, I have to go back to his LinkedIn profile because clearly, I did not realize this person actually ran an academic program. Second of all, I needed to go to this link that he sent me and become more familiar with this program to see if I think it's a good fit.

When I get to his LinkedIn profile it's there. It's like the,  top work experience. I don't know how I missed it. Director of Faculty for the Master of Science in Business Analytics program at the College of William & Mary. I'm like, "Okay. Well, if this guy thinks I'd be a good candidate, he's certainly in a position to know. He sees tons of candidates for this program."

Then, I went to the link he sent me, and I immediately remembered I had seen the webpage before. For folks who aren't familiar with William & Mary, I think William & Mary is pretty well known here on the East coast. Back in the eighties, someone wrote an article and the point of their article was to look at the defining characteristics of Ivy league schools.

Then, they took those defining characteristics and overlaid them on public schools, and they said of all the public schools out there, which ones meet  these defining characteristics of Ivy league schools. They found eight schools that checked all the boxes. One of them was the College of William & Mary.

So, when I had initially looked at that program, I immediately scratched off the list. I was like I went to this little bitty liberal arts program for my bachelor's degree that nobody's ever heard of at a private school. I don't have a strong STEM or mathematical background, which we already talked about.

They're not going to want me. So, I just struck through them and moved on. So, now , this is my first encounter with imposter syndrome. I definitely want to talk more about that in this episode because I don't think I've met a person yet in this field that doesn't deal with that. But I had this person who was in a position to know if I'd be a good candidate and he's telling me you should apply.

I'm impressed with what I see. Then, I'm looking at their program and how prestigious it is, and I'm looking at me thinking, "There's no way I can do this. I'm going to crash and burn."

I think probably one of the things that served as the most motivation for me to go ahead and at least apply was earlier in 2018, our first daughter had been born. So at this point, she's nine months old, and that had me very thoughtful about the things I wanted her to do. The stars I wanted her to shoot for , and  thought, " I don't know if I can do this, but if I don't do it, I'm going to always live with that thought of what if I had tried? And, what if I do it, and I Excel? Then, I set that bar, which is then something that my daughter can always look at and aspire towards. Then, I become a source of motivation for her, her education, and her career" So, at that point, I felt like I had to do it. I had to at least apply. Just to see what would happen. I'd already been accepted to another school. So, there was really no skin off my back if the admissions committee didn't want me.

So I did, I went ahead and applied and , I got in. Then, as I mentioned to you earlier, I just treated it like, how do you eat an elephant? One bite at a time. And, I did that with each course all the way through.

For the 18 months that I was in that program, anytime that I wasn't working sleeping, I was working on this graduate degree . But, I  graduated with a 4.0 at the end, and that was for somebody who's just literally, you look at me that process for me that we were raising the Titanic off the ocean floor. Like I started with nothing. So that's possible.

One of the things that I always wanted when I started the Veterans and Data Science Machine Learning community, and then later with this podcast, is I want to be real with people cause I feel like that's the only real way for us to combat imposter syndrome.

If you're not familiar, most everybody knows the feeling but maybe you haven't put it into words, imposter syndrome is that sensation when you're with other people who you feel like are completely knowledgeable on a subject and you feel like you're inadequately knowledgeable and that at any given moment you're going to be exposed as a fraud.

The reality of it is, especially when you're getting into data science and machine learning, nobody knows it all. You have this perception that everybody else around you has the tiger by the tail and you're the only one that doesn't. But, that's not true.

 Going back to my comment earlier about me and my situation, bringing the Titanic off the bottom. If I can get into this, learn these things, do well in grad school,  be gainfully employed doing data science and machine learning, if I can do that, then there's no reason for anybody else to have a debilitating case of imposter syndrome where they're scared to even get into this. Or, they think, I don't want to try and fail, so I'm just not going to do it. Don't cheat yourself.

If this is not something you're interested in, if you've just heard that data science is the most attractive job of the 21st century, then that's not a good reason to pursue these careers and pursue this education. But. If you've taken the time to familiarize yourself with data science and machine learning and what it is, and a little bit about how it works, and that appeals to you, and then you looked into Python and R and you learn a little bit about programming and that gets you excited. If that's the case, then don't let any other fear holds you back because you can absolutely do it. Especially as Veterans, because most of us still have some access to the GI bill or the VET TEC program. For a lot of other people, there can be very real financial barriers that would prevent them.

But as Veterans, most of us have access to financial resources that could get us the training. The only thing holding most Veterans back would be fear. Don't let fear hold you back.

Chris Sanchez: [00:25:34] No, I think you never come out and explicitly said it, Ted, but underneath everything that you've been talking about, just going from, where you were in the military to, eventually getting a career in data science, under underlying all of that is you actually had an interest.

You didn't look at the writing on the wall and say it's "Oh, the future of my field is data science. I've got to get into that. Whether I like it or not" No, it was like, you had an interest right, in learning more about statistics, in programming.

You may not have had a great background. You actually, were interested in learning math.

Ted Hallum: [00:26:12] Exactly. Yeah. I was bumping up against these concepts at work and then getting excited about it and going home and researching them in my own time in the evening. That was like when I turned off the TV and stopped wasting two hours of my life every night, watching television and started doing some productive.

That was when I bought some books. Signed up for some Coursera courses. It takes an investment, but that's where the investment started to get into this.

Chris Sanchez: [00:26:38] I think people just in general should pay attention to what they're doing in their downtime. Right? Pay attention, like what they're interested in. One thing that I can think of is, I was headed down this route to become a doctor. I was taking pre-med courses originally. It was becoming a, wanting to be a PA.

 Then I changed that along the way to, "Hey, I want to become a doctor." That was because of the, I guess peer pressure I was getting. People were like, "Oh, you don't want to be a PA go the full Monte and become a doctor", and my mom was like, "Oh, I've always wanted you to be a doctor."

So, I was like, "Oh, I should become a doctor then." I get it. I'm two years into the pre-med, and I'm taking this genetics course, and I just threw my hands up. I'm like, "this is so not interesting to me." And I realized it's I'm not going home and watching shows about ER. I'm not reading medical journals. I realized, I liked the idea of having an MD next to my name more than I liked the thought of actually practicing medicine. So, that was a red flag. I got off that path, fortunately, and continued on my own path. But, just for the listeners right now, I think that's a really important thing to pay attention to , what do you find yourself doing in your downtime?

If you're hearing about this data science thing, and you find yourself wanting to learn about Python, you want them to learn how to backpropagation algorithm works, then there's that little spark right there is enough to light  a bonfire.

 Okay, so you get through grad school real quick question for you. What were some of your favorite courses during that time?

Ted Hallum: [00:28:08] Hands down, my first machine learning course was a favorite. That course focused on teaching classic statistical learning models within the context of R. I loved that course. Then, towards the end of the program William & Mary's Master of Science in Business Analytics had a program just called artificial intelligence and it focused on teaching artificial neural networks within the context of TensorFlow with the Keras API.

After the first machine learning course , and then the second course, which focused on the artificial neural networks, I was like, for me, machine learning and deep learning, that is super exciting.  I want to get on that career trajectory where I'm doing those things to make a living.

Chris Sanchez: [00:28:50] Were you allowed to specialize in your course or was it a same track for everybody?

Ted Hallum: [00:28:55] The same track for everybody.  Actually the other program that I got accepted to at a different school did have a few tracks, but I looked at some of the tracks and some of the tracks looked really challenging and other tracks actually looked really easy. So, I was a little bit concerned about that. I thought, wow, if anybody's familiar with that program, then they'll know that it's possible to choose an easy track.

When I contrasted that with William & Mary's program, yeah, I'd only had one track. But, it was an all-star track. It had all the courses that you had machine learning one, it had a second machine learning course at an artificial neural networks course. You had to do the statistics. You had to do the advanced probability. You had to do the linear algebra. It was just a thoroughbred program where anybody who went through it, and got that stamp of approval at the end, you're legit. I felt like that program, out in industry, is going to have the kind of reputation that I wanted for it to be worth the money and the time and the effort.

I will say this when I held up William & Mary's curriculum against me and where I was starting from, I didn't have confidence I could do it.  But I thought, "If I do this program, if I can succeed, If by some miracle I don't crash and burn, then I will be able to perform at a very high level and probably won't have to deal with the same level of imposter syndrome, because there'll be a sense of accomplishment."

That's another antidote to imposter syndrome.  The first antidote is coming to the awareness that everybody else is not as omniscient, as you think they are, that they've got their Achilles heels too, and, then, the second is getting a few wins. Once you have a few wins, once there's some notches on your belt and you have a sense of accomplishment, there's no longer a compelling rationale  to think you can't do it or to think that you're a fraud.

Chris Sanchez: [00:30:51] So , I didn't have a chance to ask you , were you able to apply what you were learning to your job at hand? Some people, they're taking this data science degree and they don't really necessarily get to apply that to practical work , Which is, in my experience, invaluable . How did that work out for you?

Ted Hallum: [00:31:08] I'm glad you asked, because one of the best ways to comprehend information and have it actually soak in is to put it into practice,  and I was very fortunate.  I was in a position and that I don't think that in our particular community is extremely unique where I still had a government security clearance that I got when I was in the military.

So, after I got accepted to the College of William & Mary, I just wanted to signal to my LinkedIn network that I was going to be taking my career in a different direction. So, on my LinkedIn, up the tagline up at the top, I just changed it to "aspiring data scientist", not thinking that there would be any near term career ramifications.

Chris Sanchez: [00:31:49] Right? Because who wants to hire an aspiring data scientists, right? I want to hire a data scientist.

Ted Hallum: [00:31:53] Exactly. I was just trying to tell people I'm going to be going in a different direction,  and in my mind that meant that I would be going to grad school and that I would spend a couple of years acquiring the skills and the tools that I would need to then be qualified for an entry level data science position. But, little did I know having a clearance gave me a very unique edge in the job market because when you have companies that need to fill positions for the government, that deal with classified data, they need a layering of skills and qualifications.

They need somebody who, first of all, has the clearance, possibly - depending on the position - somebody who has a polygraph. Then, you need somebody, hopefully, that has some domain experience. That would be where background with the military background with the intelligence community comes into play. Then, finally, somebody who has the skills with data science.

So you stack those on top of one another, like in a Venn diagram, and all of a sudden there's this like tiny little spot in the middle where only a few people are applicable for those roles. I was fortunate that I fell in one company's tiny little overlapping section of the Venn diagram. They messaged me and said, we'd like to have a phone call about an entry level data science role.

I called them back, just as a courtesy, to explain that there had been a mistake . They were very polite. They let me talk for probably five or 10 minutes about why I wouldn't be a qualified candidate.

Then, when I finally shut up, they said we appreciate your honesty and the concerns that you have, but you've got to see things from our perspective. Then, they went into a description of their circumstances. They said, " We can hire somebody who just graduated from a Master of Science in Data Science program . They're going to have the data science skills. They're not going to have any domain expertise. They've never probably had touch points with the military, the defense community, or the intelligence community. Even more importantly than that, they're not going to have a security clearance. So before they can do any work applicable to the role, they're probably going to spend 12 to 18 months with a pretty high salary doing  an administrative job until their clearance could come through. But, even that is contingent upon the client being willing to sponsor them for clearance. Quite honestly, a lot of times that's a hassle, and our client just isn't interested. The alternative is to take someone like yourself who has the security credentials , has the domain experience, lots of touch points with the military and the intelligence community. And yeah, on the data science side, you're weak. But, you're honest about that. You've been accepted to a graduate school program. In about a month, you're going to start classes. Everyday, you're going to be more skilled. You're going to be able to bring those skills into the workplace, and you're going to actually be able to go to the workplace because you've got the clearance."

So, they said, "We're a small company. Our proposition is this: Why don't you come grow with us? And we'll grow with you." And I said, "Where do I sign? That sounds fantastic!" So, I actually got my first data science role, appropriately a junior data scientist role, about a month before I started graduate school.

Then, back to your original point, once I started graduate school, on a regular basis I would be fighting through a problem at work and then working on a project at home for the graduate degree in the evening, and I would learn what I needed to learn and then go back into work for the next day and fix the problem I'd had the day before.

So it was literally like sometimes a 12 hour cycle of learning a concept and then being able to put it into practice as a junior data scientist. So it really did help. It was very much a reinforcing cycle of learning and then being able to apply the same concepts at work.

That's a tremendously valuable opportunity that you were afforded there. How did you come across that job again?

They reached out to me. I just changed my LinkedIn tagline to read aspiring data scientists, and that was appealing enough that they sent me a direct message and then we scheduled a phone call.

Chris Sanchez: [00:36:12] Wow. So they reached out to you. Two points out of that story, for all the, for the listeners out there who are cursing your name right now, who're trying to find entry level data science positions and can't, right? Because it's tough. For the way that the yours turned out , I'm sure there's a lot of haters out there, but more importantly, I think is  dial-in and your LinkedIn profile, right? It's important.

Ted Hallum: [00:36:31] It is important. I think another thing worth underscoring is I totally get it. If you're coming out of the military and you've been in the Military Intelligence Corps, you've probably been cooped up for four more years in a, they call it a SCIF, secret compartmented information facility, which is a stifling environment.

You can't have any of your technology. You can't have your phone with you during the day at work. And a lot of people, as they go to exit the military, it's like the number one thing on their mind is, "I want to get out of the SCIF environment." I want to work someplace normal. I want to be able to have my iPhone in my pocket all day long.

I get it. I get that the SCIF environment is not the most exciting environment and it can be a stifling environment for data science. It's not always the most accommodating environment, but your first data science job or machine learning job is going to be your hardest to get because you're not a known commodity beyond maybe whatever you've put on GitHub.

So, every job you get after your first is going to be easier - especially if you were in your first job for more than a year or two years because people then don't feel like they're taking on as much risk. You worked there for a year or more and you didn't get fired. So you can't be too terrible.

But, when you've never had a data science role, employers look at that as they're taking a pretty substantial risk in most cases. However, in the cleared community, they're just so few people that have both the clearance and the skills, a lot of times employers are forced to make that gamble more often than not.

So, if you have a clearance and you're running away from the SCIF environment like your hair's on fire, my challenge to you would be: Consider for a year or two, at least, leveraging your clearance to get a position as a data scientist. Then, once you've gotten a year or two under your belt, and you can point to a few professional accomplishments, then start to consider uncleared work opportunities. Because, at that point, you're going to be so much more competitive. It's not impossible to come out and immediately get an entry-level job in uncleared industry, but , because you don't have any resume experience to list, you're going to have to necessarily take less money. Also there's probably going to be a little bit of attitude that comes with it of "And don't forget, we gave you your breakout opportunity. You're lucky to have this job." And, that's not really the best atmosphere to have to work within when you're first starting out.

So it's better to go someplace where you're celebrated for the credentials you do have, those things that you worked hard for in the military - hip pivot, those to your advantage. Get the year or two of data science work experience. Then you're so much better postured to go out and negotiate good salary and good benefits on the outside. If that's what you aspire to do.

Chris Sanchez: [00:39:16] You're describing what I call the "double whammy" for Veterans, which is, hey, you just spent, X number of years in the military environment. So, you don't necessarily know a lot about the the finance industry, the retail clothing industry, or healthcare industry .

So, transitioning into an industry, no matter who you are, is challenging enough. When you layer on top of that, this data science technical skill set, I call that the double whammy. Because you're, you will be fighting against people who have 10 years of experience in the retail clothing industry, and then decided to become a data scientist. And, to be quite frank, that person is nine times out of 10 and get to be more valuable in their role because they do have that domain knowledge.

Ted Hallum: [00:40:02] It's relatively easy to answer a question. The problem is, are you answering the right question? That's the real trick and people with domain knowledge know the right questions to ask, and when they know the right questions to ask, then they can provide the most appropriate answers.

That's what stakeholders are really after: Bottomline, business-affecting answers. And if you've asked the wrong questions, then automatically your answers are inapplicable.

Chris Sanchez: [00:40:27] So your first job was at this small contracting firm. And then tell us a little bit about. So your second job, was it in fact easier to get than your first?

Ted Hallum: [00:40:38] It actually came to me in a very similar way. So, I was in my first role for about a year. I loved the little company that I worked for. They were extremely supportive from day one.

But , after about a year, Battelle reached out to me through LinkedIn. One of the recruiters said, "Hey, we have a data science position at the National Ground Intelligence Center" - where I had been as an Intel guy for the better part of a decade. They said, "We know you're familiar with their mission, the human terrain, you'd be a great fit - especially since you're three-quarters of the way through your data science related degree program."

Chris Sanchez: [00:41:12] So you still hadn't even finished that?

Ted Hallum: [00:41:14] Still had not finished that.

Chris Sanchez: [00:41:15] Damn!

Ted Hallum: [00:41:16] I was working on it.

Chris Sanchez: [00:41:17] The stars just align with you didn't they!

Ted Hallum: [00:41:21] So, I said tell me a little bit more about it, and that is the downside to classified work because I'm having this conversation with a recruiter first through LinkedIn, then on the phone. There's very limited information they can give you about the actual job. That's not untypical for an interview process about a job in a classified space. So, I got very ambiguous information about it, but it sounded interesting. It was a full performance data science position. From a career standpoint, it took me on the right trajectory. The pay was a little bit better.

It was about a 15% pay increase from where I had been with my initial position. And, it leveraged again, my momentum, I talked about that a second ago. I was going back to a very familiar your environment . They also had an excellent tuition reimbursement benefit .

I had a prorated post 9-11 GI bill. So, I had 70% of the post 9-11 GI bill, and then 30% was still on me. That made the company's tuition reimbursement benefit very attractive. So, I told Battelle,  " Yeah, again, where do I sign? Sounds good." So, after a year of my first junior data science position, I roll into a full performance data science role.

I mentioned earlier that one of the things I want in this community, one of the things I want this podcast, is to just be real because being real helps other people navigate their careers. It helps give them a lens through which to interpret their own data science path experience.

So, just being real , I get onsite the first day, I get introduced to the client that I'm going to be supporting, and I said, "Okay, as your new data scientist, the first order of business for me is to find out what data you have and what insight you hope to derive from it."

And there's this long awkward pause - felt like an eternity. It was probably like five or more seconds, something like that. And finally, to this gentleman's credit, he was very candid with me. He said to be honest with you I've got a need here, where I went to every Contract Officer's Representative - they call them a COR - in this organization, and what I actually need is a geospatial Python developer. But, I went to each COR asking them about their portfolio of contracts, and did they have a position for a geospatial Python developer, and time after time I got told "No." - until, finally, I spoke to this one, COR. She said, "In my portfolio of contracts, I do have one data science contract and data scientists know Python."

So, I think that could work for you, and so they dedicated one of the data scientists billets to support this guy's mission. This was the lucky guy. He said, I didn't really want a data scientist. I need you to do geospatial Python development - and I was like, "Okay, this is not really what I expected, but here I am. I've already worked my two week notice at my other job. We're gonna roll with it."

Although the, it can be issues. They call it contracting because there's an established agreement between the company and the government about what services are going to be provided, and, if they're radically different, it's called being out of scope on a contract - which means work has been performed that's out of line with whatever the government and the company has agreed on. That can cause legal problems on both sides on the government side or on the company side.

So, I did go to my program manager just to ask him, "Hey, the client was very candid with me that they didn't want a data scientist, but I'm in a data scientist position doing this work that's really not data science work. It has nothing to do with statistics. It has nothing to do with building models."

Chris Sanchez: [00:44:55] What did it have to do with it?

Ted Hallum: [00:44:57] Well, I did Python on daily basis. So there was that. I was in that role for over a year. So, at least I got to use Python, and I did get to work with some geospatial data - but not in the way that you would normally think of a data scientist processing data. There was certainly no machine learning involved whatsoever.

So, I talked to the program manager about it and he said "Let's do this. Let's schedule you to talk with the head data scientist on the contract tomorrow. We'll see what he thinks and go from there."

The next day, I tell him about the conversation I'd have with the client. As I'm telling the story, he's just sitting across from me chuckling. He said "Ted, let me be honest with you, this contract wasn't created by data scientists or machine learning engineers. Basically, the way the contract reads, what qualifies someone as a data scientist is that they use R or Python and that they work with data. So you've told me the client wants you to use Python. Are you going to be processing any data in any way with Python? And, I was like, "Well, he says he has some geospatial data he wants me to use."

He said "The way this contract is written, that's data science." So, that was what it was, and I embarked on learning how to use both flavors of geospatial data - geospatial data can be raster format or vector format. Vector's like points, lines, polygons. Raster is a image - usually like imagery of the earth. I learned how to process that with Python in conjunction with a proprietary tool called ArcGIS. Looking back, it was actually a good experience, although it was never on my bucket list to be a geospatial Python developer. There's a lot of cool stuff you can do with geospatial data.

I would have never understood how to leverage that data and get value out of it. Now, that's another tool in my tool belt.  I'm glad I had the experience, despite that it wasn't on my bucket list. I don't think I'll ever regret being able to work with geospatial data because it's unique. It's not like working with tabular data. It's not like working with unstructured text. It's its own beast and you don't learn it overnight.

The point in that story was your career is bound and determined to take you in places that you don't expect, you have to roll with the punches, and that the upside to leveraging your clearance is you're going to be more competitive. There's less supply for that demand in the clear contracting space, but during the interview phase you have less visibility on what you're applying for.

So, you're going to have to be comfortable with assuming more risk if you're seeking a data science opportunity in the cleared environment.

Chris Sanchez: [00:47:33] When I looked at your LinkedIn profile  had actually assumed you were a geospatial analyst at some point in your military career. So that's interesting that you weren't, and that you were able to be successful in spite of that.

Ted Hallum: [00:47:46] It was very unsettling in the beginning and it brought me to another point where I'd take a deep breath, not let anxiety take control, and then just eat the eight elephant one bite at a time. Only worry about what do I have to accomplish to make the government client happy today, and then come in the next work the next day and move on to the next objective and just take it one objective at a time.

Because if I just stopped and took in the whole landscape of the tools I needed that I didn't have to be a geospatial Python developer, that would have been overwhelming. I would have been frantically looking for my next job , and you're not in a position of power to negotiate when you're leaving a job and you've only been there for three weeks. People look at you and they're like what's wrong with you? Why have you only been there three weeks and you all already want another job?

It was to my advantage to just suck it up and figure it out. But, to do that, you just have to put your head down and deal with one problem at a time.

Chris Sanchez: [00:48:44] Like a lot of things in life, it sounds like you were grateful for having gone through that experience. You were glad you had that experience, but you weren't so glad that you're now still at that job. You're somewhere new. Tell us a little bit about that .

Ted Hallum: [00:48:57] A recruiter from a company called Octo reached out to me. They saw my profile and they said, "Hey, we've got some positions that focus heavily on machine learning and you look like you'd be a good fit, specifically senior defense machine learning engineer."

I heard that and I thought, "That's fantastic!" That is exactly in line with the direction I want my career to go. The focus is on machine learning. The word "defense" is even in there. So that means once again, what am I doing? I'm taking that background that I have and the DoD and the intelligence community, and I'm going to be able to leverage that in this position.

Chris Sanchez: [00:49:27] Hits the tin ring.

Ted Hallum: [00:49:29] Exactly. Yeah, so I told them I am absolutely interested. I ended up being able to have a conversation with the hiring manager, that conversation went well, they made me an offer, and I came on board with them in January .

It's awesome environment. They support the DoD, but they themselves are a tech company.

So, I'm on a team that's like a R and D team. I'm helping to create new services that are powered by machine learning that will be made available to government clients.

Chris Sanchez: [00:49:58] It sounds like that's where you want to be, and you're actually developing some of the pipelines for putting machine learning models into production for the DoD.   

Ted Hallum: [00:50:08] That  right, and that's very exciting for me too, because not only is it the direction I wanted to go with machine learning, but it's like like cutting edge.

Building a model is one thing, but if you really want to be serious about a model, you can't just build it. Building it as the easy part. The hard part is taking it and putting it into production in a way that's both scalable and sustainable over time. So, as data drifts and things change, your model performance is probably going to go down . You need to have an automated, sustainable way to take that model and put it back through the retraining process and keep it performing optimally.

So, I'm part of a team that's putting together a pipeline like that, which is really exciting because there are lots of places that just have their model deployed on a little Rest API or microservice type thing, but they don't have any process in place to maintain that model over time.

More and more companies are taking a cue from the software development community and the whole DevOps process. There's a modified version called MLOps. That's the terminology for this pipeline that I'm talking about. So, I feel very fortunate that my career has taken me to a place where I can get experience with MLOps. 

Chris Sanchez: [00:51:19] All right, so we started in Georgia, got to 9-11, military, series of contacting jobs after that, breaking into data science, College of William & Mary, first data science job, second data science job, and career Nirvana where you are right now. Where does the Veterans in Data Science and Machine Learning community get involved in that picture and The Data Canteen for that matter?

Ted Hallum: [00:51:48] So when I started off,  I felt like I was in a dark room looking for a light switch all by myself. And, unfortunately, there was no group at the time, especially not for Veterans, where you could go and get advice, mentorship, hear about other people's experience , have camaraderie in the field. There just wasn't anything like that.

Of course, when I was in grad school, I didn't have enough knowledge to really be helpful to anybody at that point. I was also overwhelmed because I was in classes. But, it had been in the back of my mind that at some point in the future, I would want to try to create a community to provide those things to other people who are going to come along that same path behind me .

 So, it was November of last year,  I'm far enough in this journey where  I know enough to be useful to other people who want to do this. I get to Veteran's Day , and I'm scrolling through my LinkedIn feed. Of course, they're all the typical Veteran related posts because it's Veteran's day, and it's this thought comes back to me about this community,  and I'm like "If I'm going to do this, I need to pull the trigger today. It's going to get more traction today than any other of the next 364 days."

And I'm like, "I'm going to do this as a LinkedIn community. That's how we'll get started." Then, I just started inviting people from my personal LinkedIn network  and I made it open so that they could then in-turn invite people  who are both in data science and machine learning and who share in that common bond of U.S. Military service.

Inherently, by the nature of  having to be a Veteran and having to be in this field, it's never going to go viral. There's never going to be millions of us, but I started thinking it would be cool if I could find 30 or 40 other people and here we are going on having 500 people in the community. I'm ecstatic. I feel like it's gone gangbusters.

It wasn't too long after it got started that somebody said, we should have a podcast that would really serve the needs of the community. At that point, I took a deep breath cause I was like there's a big difference between starting a LinkedIn community, which I was able to do in two hours time, and and starting a podcast - which I had no experience with. And, again, imposter syndrome was the ever-present Wolf at the door. Cause I was like, if I'm going to have a podcast and I have to be "an authority." People are going to be nitpicking every little thing I say! But, I saw the validity in the recommendation, and I thought " if I really want to  establish a comradery between folks, then having a podcast and bringing members of the community on the show, that's certainly a great way to do it."

I've had people reach out to me already. We've only done a few episodes, and people have sent me direct messages. I've gotten emails about how they weren't sure what to do, they listened to an episode, and now they have clarity and they know exactly how they need to proceed.

And. I'm like, boom, that's the point! That's what all this is for. I feel like if it can do that for a few people already , when we're in our relative infancy, I think we're gonna be able to help a lot more people, and that gets me excited!

Chris Sanchez: [00:54:43] Love the mission, love what you're doing. and I think a lot of people probably feel the same way that you do - that they want to give back in some way or create a community of like-minded people . But the difference between you and most is that you actually took action and built it,  from scratch. So there's something to be said for that.

Ted Hallum: [00:55:03] There's that saying "sometimes you just have to be the change you seek." I didn't see a place for people like us. So, I just felt like there was no alternative. If it was ever going to happen, maybe it would just have to be me. So I tried to be that change.

For people who might be interested in, maybe you didn't know about the Veterans in Data Science & Machine Learning community, and you just stumbled on this podcast.

I do want to throw up our URL. You can go to the website and learn a lot more. Of course, there's the podcast here, which will be a biweekly benefit to you.  In addition to that, a big focus for us is one-on-one mentorship. So we actually have people in our community, some of which were in data science a decade before data science was a term, and they're willing to take members of our community who want to transition from the military , or who are just Veterans that want to transition into data science and machine learning, they're willing to take them under their wing and provide one-on-one mentorship.

That takes what you get here in the podcast, or what you get in the LinkedIn community, and put an exponent on it. That's the kind of benefit you're going to get from having that one-on-one connection. If you're interested in being a mentor, or if you could benefit from having a mentor, I would recommend you go to this URL and that's where you can sign up to be a mentor or  request to be a mentee . There's no cost associated with any of this stuff. We are a community by Veterans for Veterans who want to help one another . To help cover the cost of hosting the website, the podcast, and all that you can make a donation.

In the future, we may pursue having a Patrion program where you can pay a little bit of money and get access to extra stuff, but the general having a community to belong to, having these podcasts to get advice from, that is always going to be available to any Veteran and there won't be any cost associated with it.  Again, the mentorship is another thing, no cost associated with that. Just come, be honest about needing help, and we'll try to get you paired up with someone appropriate both in terms of simple things like time zone and who actually has an interest in the vertical of data science that you're interested in.

Chris Sanchez: [00:56:57] Thank you for expounding on some of those deep thought-provoking questions that I asked you. I got a few softballs to toss your way.

Ted Hallum: [00:57:06] Alright, let's do it!

Chris Sanchez: [00:57:08] I'll start with what is your favorite programming language and why?

Ted Hallum: [00:57:12] My favorite programming language is Python. There are a lot of reasons for that, but the number one reason I think is pandas. There's so much time and effort that has to be expended on data wrangling, and I just don't know of a more user-friendly way to get through that. I would almost want to see pandas get ported over to any other language that would seriously use on a day-to-day basis.

Chris Sanchez: [00:57:38] Do you have experience with R?

Ted Hallum: [00:57:40] I do. I used R in the Master of Science in Business Analytics program at the College of William & Mary, and I liked R. R also has data frames.  They operate a little bit different.

One thing I will say about R is for the classic statistical learning approaches, I really liked the caret interface. I don't know if you've ever used that. It's a package for machine learning in R, and the goal of it is basically to be a unifying interface for all of the different classic statistical learning implementations that are out there. I was pretty impressed. It works well. So I like that about R. There are other things I like about R . It can be a bit more natural for things to be indexed based on one, rather than indexed based on zero like in Python. But, once you get used to things being zero indexed, that works out fine too.

Chris Sanchez: [00:58:27] Speaking about pandas, I was always blown away with the story of West McKinney.  The creator of pandas, quant analyst and the tools he had just weren't getting done. So, what does he do? He builds his own. He builds pandas. It's a well curated ecosystem now. It's got a whole community developers behind it, but I'm just blown away that, someone just took the bull by the horns and created something that is literally used by people worldwide today.

Ted Hallum: [00:58:50] Absolutely! It's been designed and evolved in such a way that it pairs so well with other essential things, like for instance, SQLAlchemy . I haven't found a better way to get data into a database or out of a database than to take pandas and pair it up with SQLAlchemy and let those two do the heavy lifting . It is just phenomenal.

Okay. So what other data science or machine learning technologies get you excited?

 If you want to talk about technologies related to artificial networks, I love TensorFlow and Keras. I think that's a great combination. I find it to be very user-friendly. I've dabbled with a couple other frameworks, and I'm sure I'll use other frameworks in the future, but , for the day to day, if I was just going to sit down and hammer out a prototype, TensorFlow and Keras would be my go-to tools in concert with Python.

Chris Sanchez: [00:59:42] Any experience with, PyTorch?

Ted Hallum: [00:59:44] In a MOOC! I did a machine learning track on DataCamp, and I believe it was one of their modules in the machine learning track.  From that brief exposure to it, which may not have been the best introduction,  it just wasn't as user-friendly. I remember going through the exercise and I thought, "Wow , this is a bit more cumbersome than it would be with TensorFlow and Keras."

Chris Sanchez: [01:00:07] I honestly think that speaks to how far the Keras and TensorFlow packages have come.

I remember back of the day, writing it in 2017 , very different from writing it today - a lot less user-friendly.

Ted Hallum: [01:00:23] From what I've read, that's been a focus specifically on the Keras side.  Usability has been something that they've really poured their hearts into trying to make that better than what it once was.

Chris Sanchez: [01:00:34] Yeah, and I can vouch that they have. Okay. So speaking of MOOCs, what's the next one on your list?

Ted Hallum: [01:00:41] The next MOOC for me is going to be a course called "Self-driving Cars with Ducky Town." That's actually a collaboration between the University of Montreal, the Toyota Technical Institute in Chicago, and the Ducky Town Foundation. They've come together to offer this course - that's now going to start in March. It was originally going to start in February.

You can do the course in conjunction with a kit that's based on the Jetson Nano board. Throughout the course, you're going to take machine learning and combine it with robotics, so your little self-driving car, and you have to implement classic procedures in conjunction with machine learning to have this little card and navigate the course, dodge pedestrians, and not go off the road. I haven't researched this claim, but they claim it's the first MOOC ever to teach a scaled-down implementation for a self-driving car.

It's supposed to run for 10 weeks, and you do it with Python and TensorFlow. Then, I believe the models get deployed to the little Jetson Nano board with Docker, which I'm looking forward to.

Chris Sanchez: [01:01:52] Not asynchronous mean not self-paced you're doing it with a cohort.

Ted Hallum: [01:01:56] That's right. Yeah.

They may do future offerings, but for right now, there's only one offering that's scheduled. I believe the new date is March 22nd and it will run for 10 weeks.

Chris Sanchez: [01:02:05] Now, last I checked, you didn't work or you don't work at Tesla or Waymo. So why self driving cars?

Ted Hallum: [01:02:12] I do find self-driving cars to be cool, but I wasn't out to find a course that would teach me about self-driving cars. I more wanted to learn about how to take my machine learning skills and implement them in concert with robotics.

That's what this will do, and it's super cool that it's based on a Jetson Nano . I'll be implementing my models on a legit edge device, which I think that there's going to be more and more of that happening in the future.  There's a lot of little boxes that I want to tick while it goes through the course, like what's the proper way to deploy a model to an edge device. You know, I'll learn that. I've never done that.

Then, I would think that there's considerations that you have to make about the types of model that you can implement and how to get and a good balance between performance and efficiency because it's not an AWS server. There's a difference in the amount of compute that you can have this little self-driving car do, but yet it still has to make inference that's good enough so it's not running over the faux pedestrians and going off the the course that you've setup for it to navigate.  I see a lot of potential. Of course it'll all come down to the actual content and how it's executed, but I'm excited to see what they bring to the table.

Chris Sanchez: [01:03:21] It definitely sounds like a more advanced MOOC because it's going to combine a bunch of different skillsets and put them all together. So, if you don't have intermediate level Python to start with, it doesn't sound like this would be the course for you.

Ted Hallum: [01:03:34] I would suspect the intermediate knowledge of Python is probably a pretty comfortable place to start the course.

Chris Sanchez: [01:03:40] Is this that your audience can still get into if they're interested?

Ted Hallum: [01:03:45] I think you can. Like I said, the date has been pushed to March 22nd. I don't know what their new timeline is for people who are just ordering the little Jetson Nano- based kit. You may still be able to get that before the class starts on March 22nd. But, worst case, they do have a track of the course where you can just build your model and deploy it in a simulation.

So, you should still be able to get all the fundamental benefits of the course, even if you don't have a kit on the day that the course starts.

Chris Sanchez: [01:04:14] I'm sure you'll put the link in the show notes.

Ted Hallum: [01:04:17] Absolutely.

Chris Sanchez: [01:04:18] To wrap things up, what do you have in terms of recommendations for any kind of data related podcasts, books, or learning materials for the interested audience?

Ted Hallum: [01:04:29] There are so many great ones. I wanted to cherry pick a few that I think will be good for the spectrum of our listeners. For those who are just getting into data science, I wanted to recommend one of the podcasts that served as a inspiration to me.

It's still available. They're no longer producing new episodes, but you can go back and listen to all the historical episodes. The host is a data scientist named Renee Teate, and the podcast na m e is "Becoming a Data Scientist". She just brought on one guest after another, and let that person talk about their path. How did they go from whatever it was that they used to be to the data scientist or machine learning engineer that they are today? Hearing those stories, people from all different backgrounds, was hugely beneficial to me. Particularly people that had a background that I could resonate with. It made me feel like I had a template that I could follow. So those are still very relevant.

The second one is a podcast called Eye on AI by Craig Smith. He is a former correspondent for the New York times, and this podcast is different. He doesn't generally bring on data scientists to talk about how they became a data scientist. His podcast is very strategic in nature. The implications of machine learning, deep learning, and artificial neural networks at the national and global scale. He's got heavy involvement with the National Security Council on Artificial Intelligence , and they're starting to make draft recommendations to Congress  just within the last few weeks actually. So that's a super informative podcast for more of that high altitude perspective on how machine learning and artificial intelligence are quite literally changing our world.  That's a unique perspective that you don't get from your day to day  hands on keyboard with the code.

I was going to stop with those two, but I recently discovered a podcast by a gentleman named Jon Krohn, the SuperDataScience podcast. I was doing some searches related to MLOps and his podcast came up because he had a guest on his show, Erica Green, she's a manager of machine learning engineers at Etsy. Excellent episode with her about the stuff that she's doing with machine learning and MLOps. I immediately liked the podcast after I stumbled on that episode.

As far as books, I've mentioned this book on the podcast before, I would be remiss if I didn't mention it again: The book called "Build a Data Science Career" by Jacqueline Nolis and Emily Robinson. As far as the one-stop shop , from I've heard the phrase data science all the way to I'm two months into my first data science job and trying to be successful, this is the one book. I have  not seen any other books like it. The breadth that they cover for people who are getting started in data science and machine learning is just incredible. They do a really good job.

The second book I would recommend is a book that was on my personal reading this back in 2019, and I still think about this book on a pretty regular basis. It's called "AI Superpowers: China, Silicon Valley, and the New World Order." It's written by Kai-Fu Lee, and, if you haven't heard of Kai-Fu Lee, he's had an incredibly storied career with the big hitters in the business -  like five or six years he was with Microsoft. Then, he was the VP of Engineering for Google for about five years. Now, he's got his own venture capital company.   He gives a very unique perspective because  he's done business in both the West and in Asia. So, he can speak very articulately to the differences in the way that AI is shaping business and culture here in the United States, as opposed to in China specifically. It's very different. A lot of people have been accustomed to the comfortable position that we've enjoyed here in the United States as being the world's dominant superpower, but he lays out a very cogent case that things are changing fast.

By definition our community, the people who listen to this podcast are Veterans, patriotic folks who want to see the United States continue to be the world's dominant superpower. If you're not abreast of how differently we in the West are approaching artificial intelligence, educating our children to be AI literate, all these dynamics compared to China, then  you need to read this book, and you don't even have to read it. There's a great Audible version. I actually consumed a large portion of this book via Audible works well in that format, you can just listen to it on your commute or during your workout or whatever. It's super informative and you're not going to get that perspective anywhere else.

Chris Sanchez: [01:09:09] You've piqued my interest. I'm definitely going to have to put it on my reading list.

Ted Hallum: [01:09:13] It'll be in the show notes.

Chris Sanchez: [01:09:15] Okay. Ted, this has been a real pleasure and an honor to be able to interview you.

What do you recommend is the best way to contact you?

Ted Hallum: [01:09:23] You can always hit me up on LinkedIn.

My username is right there on the screen. My email address will also be in the show notes. So, if you want to reach out to me via email, that's perfectly fine as well, and very soon we will be announcing our new Slack community that will be available to people who are Patreon subscribers.

 That'll be another great way to hit me up. Inside the Slack community, you can do zoom-style video calls. It adds a really neat dynamic that we don't get through LinkedIn. So, if you want to level-up your Veterans in Data Science and Machine Learning membership, that would be a great way to do it, and you can hit me up in the Slack community.

Chris Sanchez: [01:10:01] Awesome. Thank you, Ted. And thanks for having me on to have you on.

Ted Hallum: [01:10:04] Chris, I really appreciate your time. You've been so generous coming on to introduce me and put me in the hot seat so I can understand what people go through as they prepare to come on and be my guest on the show.

I couldn't be more appreciative. Thank you.

Chris Sanchez: [01:10:17] Yep. My pleasure.

Ted Hallum: [01:10:18] As always, until the next episode, a bid you clean data, low p-values, and Godspeed on your data journey.