Data Science Leaders | Episode 05 | 33:37 | June 01, 2021
Bridging the Gap Between Data Science and Business Outcomes
Satyam Priyadarshy, Managing Director for India Center, Technology Fellow, and Chief Data Scientist
Halliburton
Data Science Leaders | Episode 05 | 33:37 | June 01, 2021
Satyam Priyadarshy, Managing Director for India Center, Technology Fellow, and Chief Data Scientist
Halliburton
The best data scientists are continually learning something new, taking on unfamiliar projects, and keeping their skills fresh.
The best leaders in the industry create a culture where teams have opportunities to grow and are able to clearly understand and communicate data science concepts.
In this episode, Dave Cole is joined by Dr. Satyam Priyadarshy, Managing Director for India Center, Technology Fellow, and Chief Data Scientist at Halliburton, to discuss how to explain complex subjects in simple terms for better business outcomes.
Dr. Priyadarshy also explained:
Hello, my name is Dave Cole and I am your host of the Data Science Leaders podcast. So let’s dive into the podcast, Dr. Satyam Priyadarshy. Mr. Priyadarshy here, is the Managing Director, Technology Fellow, and Chief Data Scientist at Halliburton. He has about 30 years of experience, Satyam, is that about right, at least, in the data science realm? Also, what’s interesting is Dr. Priyadarshy has also been a professor at both George Mason and Georgetown, and most recently at Virginia Tech, and a host of others. And we’ll talk a bit about his role as a professor, as well as what he does for his day job at Halliburton, how he balances the two and what he does there.
So starting there, the one thing I want to talk just a little bit about is, given your background in education, how do you feel that you’re able to have that balance between being a professor and taking some of that teaching and coaching ability, how has that influenced you as a leader?
Certainly. So one little correction, I’m a Managing Director of India Center for Halliburton, not the whole Halliburton, because that would be a much bigger organization. So I run the data science center and a small center in Bangalore, and, of course, a Technology Fellow and Chief Data Scientist. I’m based out of the US, but I operate the India Center from here, and that’s been an additional responsibility as of last year.
Now, coming to being an academic, I think it’s a wonderful experience and a knowledge gaining place, being in academics. Typically, people leave academics, go to industry, and just build a career there. For some reason, I could not do that. I was a full-time faculty, and then I switched from being a computational quantum chemist to being a technologist, and then eventually becoming an executive to turn around some companies.
But in the whole process, after I did my MBA, I decided enough of teaching science and computers. Let me go teach some management, and that’s how I ended up teaching at the school of business at George Mason University, where I was an adjunct for about seven years. And I taught courses on managing information systems, and things like that. I have no degree in computer science. I don’t even have a course on computer programming in my record. I’m a physical chemist, turned into a quantum chemist, and worked in various fields of computational science.
And then digital marketing, it was since I became an executive and the head of analytics. I was looking at all the marketing stuff, and that’s when I started teaching digital marketing strategy at Georgetown. And later on, today, I teach at the business school of Virginia Tech, and I’m an adjunct faculty at the computer science department of Oklahoma State as well, but also in some university in India and a visiting lecturer at too many places. One-off lectures here and there, a lot of boot camps, so to say, but that’s all for fun.
The reason why I do that is for two reasons. One is, when I left full-time academics, my parents and my PhD advisor were very unhappy, sort of. How can you leave an academic job?
Right.
But you have to do things that are good for your family and overall. So I decided that I’ll keep the job as faculty, and that makes my family happy, like my father especially. But also, it’s a passion. The main thing is, you learn a lot from younger generation, or the students, because you would think that you know something, but when they ask you the question, that makes you think very differently. A simple term could be really challenging to explain, and that is the whole game of data science, if you think of it. Data science is not about building a histogram and talking, “Okay, the sales are going north and south…” In most American, if you look at the marketing presentations, they will say “north and south,” but it’s a numerical number, whether it’s going up 10% or 12%, but we call it north and south, which is a business speak, which is fine.
But if you were to explain it to somebody and a student doesn’t understand what north and south is, then you have to explain to them in a different way. And so, simple concepts, multiple ways to explain it, is the art of data science.
Yeah. I’m glad you bucked to the trend. There’s this awful phrase that those who can’t do teach. So you’re doing both, you’re teaching and you’re also leading a team. How has that ability to talk to your students and to teach them concepts, how have you applied that to your role as a leader? Do you find yourself getting into teaching mode when you’re talking to your business counterparts?
Absolutely. So I’d give you one example, which still fascinates me. This is roughly about 15-20 years ago. I was in one of the internet companies that are giants, and we were looking at competitive intelligence data, which we generated. And in order to simplify that chart, I made what is called a box and whisker diagram. And everybody in high school learns about it. I shared it with my senior vice president at that time, and he says, “How do I interpret this?”
Yeah.
I just said him that, “You probably studied this in high school.”
But you forgot, yeah.
I had to actually create a slide, how to interpret slides.
Wow. Yeah.
And that’s the fascinating part because, if you look at it, well, there is a reason for it. Most executives, they don’t have enough time to really go spend time as a scientist would do it. And hence, you have to really explain to them in 30 seconds, if you want to call it that number.
Right.
It is within that attention span where they can digest things. As I teach my team that, whenever you create slides, the slide should be like kindergarten; four words, five words, but the story should be there.
Right. And we don’t mean to put down our executive counterparts. They’re vital. But I hear what you’re saying because, what I found, too, is… And I’m curious if you see the same thing is, when you’re forced to boil something down, when you’re forced to simplify it down to a simple takeaway, you actually find that you act… and you have to explain it. That also can be a way in which you get deeper into the analysis to really boil it down. Have you found that to be a helpful technique with your data science leaders?
Absolutely. If you can’t explain things in very simple terms, that means you have difficulty understanding it. This is a concept most physicists already know. The world is four dimensional, five dimensional, six dimensional, whatever you want to call it, three. We can’t explain two dimensions very well, so what they do is they will say, “Let’s solve a problem in one dimension fast.”
Right. Yeah.
And then, it’s much easier to explain. Then you can say, “In one dimension, this is an exact proof. In two dimensions, this is a proof. And in the third dimension, this could be the proof.” Right?
Right.
But it’s a complex problem that you boil down to a simple dimension. And in the business world, the simple dimension is how simply can you give your story, which will make decision-making much faster and more effective. That is an art and a skill that one has to learn. And it works very well. And I think this teaching ability is really very powerful.
That’s great. I can see, given your background, you, participating as an adjunct professor at a number of different universities, that’s certainly helped you. How do you impart that onto your team, themselves? Certainly, a slide for a slide is a great takeaway and teaching to a kindergartener. But is there a training that you can go? Because, sometimes, you can have data scientists who are just absolutely phenomenal data scientists, but when you ask them to explain their work, they fall all over themselves.
Yeah. Because I think it’s an organizational challenge, as I would call it. Almost all the data scientists who work in my Center of Excellence and in the oil and gas industry, almost all of them actually teach, whether in university or internally. Because when I started this, and especially in the oil and gas industry, 2014, if you think of it, that’s when I started, and I started building a team…explanation was difficult.
First of all, the first explanation was what data science will do to us because if you talk to most oil and gas leaders, they will tell you, “Oh, we have been using data for a long time. And we have the most, well, we have the largest volume of data in petabytes.” All those things are true only to an extent, but its application, its value creation has been a challenge.
One of the examples I give, the industry talked about what’s called digital oil fields, 25 years ago, roughly. And there is not a single digital oil field, right?
Right.
It’s not just having the concept. It is actually connecting and creating value. So when we build this data science team, the question was, how do I explain to people what we are doing and how it’s valuable?
And as a result, we saw that a lot of domain knowledge is needed. Data scientists have the data science concepts very well, but no domain knowledge. So if you want to communicate with the domain experts, you have to understand that bridge between data science and the domain. And as a result, most of my data scientists became, you can call it, a mentor and a coach. We actually built industry’s first series of boot camps to train people across the industry. Over the last, I would say, four years, we may have trained about a thousand people around the world in many companies.
Wow.
These are all practicing data scientists who actually build models for living, write the results, write a story. One of the things that most of them know, that when you write the slides, the story is like a movie script.
That’s great. That’s great advice.
If you think of a movie script, it’s very specific, right? It’s not like a blabber going on for too long. Every sentence has a meaning, its impact, its value, so you write the script. Even if somebody, you are not in front of the audience or the internet connection is down, if they were to read that movie story, they could say that’s a full storyline.
As a result, when they’re practicing data scientists, and then they do the teaching, it is much more powerful. And hence, we have been very successful. In fact, that was not even our revenue stream, and now, we have built that into a revenue stream. But the most important part of this was that we were doing these boot camps in a highly contextualized manner. So one can actually go take a data science course on one of the MOOCs, basically. But what we found is that people can’t translate that into reality.
Right.
A very few percentage. We all know that percentage is such a small percentage. Millions of people join a course, but how many actually complete and take it forward? That percentage is very low, historically. In fact, I really try to put my data scientists into one of these boot camps within the first six months of their work.
In these boot camps, what is the specific output that you love to see? Maybe it’s all the above, but is it the ability to take your data science work and turn it into a graphical presentation, like some visualization? Is it actually writing a document that you would hand off to your business counterparts? Or is it a deck, like a slide deck, or it could take all forms? But generally speaking, what is the form of the output of your team?
So I’d like to go a little bit deeper into that. One of the things that we developed over the last six years is a concept called “smart digital.” We publish those things, actually, in articles. But the concept is that, how do you take somebody from one point in the journey to the end of the journey, so to say?
Right.
And it’s not about just giving them, okay, here is an algorithm, or here is software that you want to do. It is actually trying to understand what business problem we are talking about, right?
Yes.
How do you come up with which problem you will be solving? And so, this whole concept is built as a smart digital workshop. And so, most of the people that we train internally, it’s to become an expert in that smart digital workshop. So that when they’re talking to, since we are a service company and we talk to customers, when they talk, they can talk intelligently about it. Because it’s not a question of, “Oh, you want to do a project?” It’s not useful. The question is, “What am I going to do which will help the business?”
Right, what business problem am I solving?
We are not a full academic institute that you can write papers and nothing to do with it. But here, in business, everything is tied to dollars and cents.
Yeah, absolutely.
As a result, we take people; we mature them and monitor them as a cohort, so they do actual projects and actually deliver the results. And then, whether with or without a partner, or they will actually come up with a full project plan as to, “Okay, we want to solve a specific problem, and this is how we will approach it. And this is how the results will look.” And in three months, six months, they will actually present it to the cohort.
Will you also talk about ROI in that, like this is the expected ROI if we’re able to increase the accuracy of the model, what have you?
Yeah. There are about five or six different financial metrics that we can look at: accuracy, efficiency, NPV, things like that.
Yes.
So any project that is taught out has to have one of those financial metrics associated with it.
I totally agree. Going back to one of the things that you said about petabytes of data, I chuckle from time to time when I hear that from data science leaders. It’s not about that. What I’m hearing from you is, it’s about the insights, it’s about the business value you’re driving. You can have all the data in the world, but if you’re not deriving insights and you’re not turning that into changes in the decision-making process or improved predictability and forecasting, or whatever, then it’s not really worth much. So that’s very, very interesting.
I’d like to segue a little bit. One of the things that, talking about teaching and talking about some of the programs and the boot camps that you’ve created, I know also you have done a phenomenal job in increasing diversity amongst your data science team. I have some stats; I’ll just throw out here for our listeners here, but this is a 2018 BCG research report. And 55% of university graduates, I assume this is around the world, are women; 35% of women have STEM degrees. So STEM, standing for science, technology, engineering, and math. And then, 25% of the STEM workforce is female, and then 15% to 22%, I’ve seen 18%, I’ve seen in that range, of data science professionals are women. So in the data science community and, I’ll speak as a representative in the science community, we’re doing an awful job, really of creating a diverse workplace when it comes to data science. But you, however, have figured some things out here. Would you like to talk a little bit about that?
Since I’ve been teaching for quite some time, as well as actually building data science or analytics or data monetization teams for a long time, there are some inherent talent when it comes to pattern recognition, and diversity really helps there. As a result, in fact, when I started building the team, one of my first hires was a PhD in atomic physics with astrophysics background, things like that. She has not worked in the oil and gas industry before. If you look at what the job descriptions are, like most people put out for data scientists, they want a rockstar in Hadoop, and programming, and machine learning, and AI. I’m not looking for such people. I’m looking for people who can actually do problem solving and can bring some diversity of thought in looking at the patterns that we will be creating from the data. And as a result of the team I built, we had people with astrophysics background, we had from MBA, we had chemical engineering, computer science, and things like that, different fields. Irrespective of their gender, the field itself was diverse.
Right.
As I’ve said, my background is quantum mechanics or quantum chemistry. I have worked in various fields. I never got blocked because of that. I got blocked for different reasons, but at least, I could do things in the technology world or in an executive suite with that degree. So when I rebuilt the team, we were very focused on how do we bring women into the fold, and we were very successful. At one point, we were at 50%. And recently, since I took over the India Center in Bangalore, I have just made offers of 66% women with master’s in computer science.
Wow.
I’ve hired people from all around India. I’m looking for the thought process. I’m looking for openness to learn new things. And everything else, you can actually train people.
That’s interesting. There’s certainly different schools of thought in terms of, should you be hiring people who have that diverse background, and don’t come out of college with a specific data science degree, or a strong statistical background? Obviously, you need to have some grounding in either coding, or math, or problem solving in general. And it sounds to me like you are looking at folks coming out of college who do have diverse majors, if you will. And I assume, you have to train them. At some point, you have to train them on data science. So, how have you been accomplishing that?
A couple of things. One is, even for the senior people that I’ve hired, experienced people who are, say, chemical engineers, and they have done quite a bit of work in the industry or outside the industry; what I’m looking for is what their problem solving skills are, whatever programming they may have done. And are they open to learning new programming techniques? We don’t want to tie them down to just use Python, or R, or something like that. That’s not the criteria. In fact, many of my hires have been literally fresh out of college, even in the US, and talking to them, “What have you done? What project have you done? How have you approached it?” And they turn out to be fantastic employees and team workers. But the main important thing is they should know how to teach others as well.
I love that.
When we hire the so-called early-career, they are always paired with some senior people. And very quickly, they pick up, and then this boot camp methodology of teaching others, works automatically in favor of me. So I reduce my productivity’s lifecycle to a very short time range. And people are more productive that way, and they’re interested. The thing is, if you give them a monotonous job, “Okay, go run neural networks every day,” I can guarantee you, most data scientists will get bored.
Right, right. I’ve always had the philosophy: if you’re not learning, you’re leaving. Clearly, if you have a culture of coaching, if you have a culture of learning, and you have an expectation that you also have to pay it forward. Not just, should you be learning something new and maybe taking on a different project? Maybe your last project was building out a deep learning classifier, and then the next project might be a forecasting project. So keeping those skills fresh, I think, is very important. But then, also creating a culture of them becoming teachers. So I imagine, if I had to guess, some of your students then become teachers and mentors as they grow in your organization. Is that assumption correct?
Absolutely. As I’ve said, we are running global boot camps, so I’m not everywhere. Some of my team members are leading those things. Somebody’s in Japan. Somebody’s in Australia. Somebody’s in India, in Indonesia. So I’m not there anymore. The team members who work with me lead this and their own sessions, right?
Right.
I don’t have to be involved day-to-day. And then, for six years, I can tell you that it has been a great rewarding journey for us in that sense. And they figured out their own… depending on what context they are trying to build the bootcamp, they can figure out, now they know the methodology. They know how to deliver. And cultural differences from Japan to Europe, to Asia, South Asia, to America is very different. They learn how to deliver in these areas.
Yeah. You have a globally distributed team, right? You just mentioned that you can’t be everywhere, micromanaging all the content that’s being produced. But is there some governance that you’re putting in place to ensure that the methodology is to your best practices? Do you worry at all about some mentor or some leader in Japan creating a siloed process?
No. That part, since I’m part of the Center of Excellence, everything boils down to that. But there is no need to micromanage. I think the trust in people is very important, and I personally believe that the team that I have, I trust them. And so, I don’t see that as a challenge.
Yeah.
They should have independent thinking to do that. But at the same time, they are fully aware of what’s within the governance framework that exists at the highest level of the business. We will stay within that. For example, one of the key concepts in oil and gas is you don’t share any data, right?
Right, right, right.
So everybody understands and people do not do such a thing. So I think that’s what it is. Otherwise, I never believed in micromanagement because it doesn’t work.
Yeah. I think there’s very few who believe in micromanagement. I think sometimes people don’t realize that they are micromanaging, so you always have to be self-aware, like that. I’m glad you’re in that camp, too. We’re switching gears, one thing I want to throw out here is, I love the time machine-type questions. So if you were to go back, say, 10, 15 years, maybe earlier in your career, what would you tell your 15-year-ago self based on what you’ve learned today?
I grew up in India and I came with a certain philosophy in mind, and I still keep to that philosophy, basically. And that philosophy is that you have to do what do you want to achieve, don’t worry about others. What do you want to achieve, you figure out a path to achieve it because nobody will take you there. You have to take yourself, right? When I left academics and joined the corporate world, very quickly, I learned that there happens to be what is called a corporate structure, and things like that. And that’s when I said, “Oh my god.” And then I learned, “Oh, there is something called a CEO.” Then, within six months of that job, I decided that I want to be in the C-suite in 10 years time because I calculated based on books you read, at a certain age, you have to do these things. And I had already spent so much time in academics. But the day I joined reporting to a CEO and a chairman, that was nine and a half years into my 10-year plan.
Nice.
So I planned everything, whether it’s teaching, whether it’s learning new skills, whatever it is, because I didn’t have any godfather in America. In fact, I was told that, “You can never become a leader in this country.”
Oh jeez.
You can figure out your path by yourself, but at the same time, you should have enough freedom as I give to all my team members. They should really figure out the path. We only show them how to achieve it, but they can actually build their own path.
That’s a lot of good thoughts in there. First of all, you clearly persevered. Coming into this country and facing some naysayers who said, “Oh, you can only get so far.” And being able to, obviously, be phenomenally successful and be the C-suite type executive as well. The second thing that you’re saying is you really have to take ownership for your destiny. And my guess is, in that 10-year plan, there’s a lot of hard work that went in on your behalf to make it to where you ended up after those 10 years. And I think, if any of our guests were to go and Google you, and look at your LinkedIn profile, it is quite impressive. Not just the teaching aspect, but the roles that you’ve had and the accomplishments that you’ve made. So I’m very excited to hear that success story.
Speaking of mentors, one other question I’m just curious about, for those data science leaders out there, or aspiring DSLs, have you had a mentor in your career that has helped you along the way? You mentioned that you really got to do it on your own and you didn’t have a godfather. But was there anybody out there who you leaned on, or did you really go it alone?
Those who have heard me, or when I speak on the stage, I have a slide for that, actually. Nothing happens without that, and I think contribution, as I say, after my parents is my teachers from my primary school to the PhD, and especially my PhD advisor and two post-doc advisors that I had. They’re all there. They had a big role in that because, as I’ve said, I’m a physical chemistry graduate, and then I did my PhD in quantum mechanics, which is a field that a very few people want to enter. He was a genius. I did my PhD in India, but he did his PhD at the University of Virginia. And he did his PhD in two and a half years at UVA and went to India. So you can imagine, it’s a very difficult field. The way he trained without pressure, without micromanaging as to how to think of problem solving. And one of the things that he did early on after two and a half years of PhD, he said, “You have to publish a paper in a different field than the PhD topic,” which is rare in many cases. That basically wired me to think of multiple problems at the same time.
Yeah.
So I owe a lot of my success to my PhD advisor. He is Professor S N Datta at IIT, Bombay. Amazing personality, genius. But at the same time knew, without shaping you, shaped you in a way that you can think of most complex problems and not to be worried about it. And then, my two post-doc advisors, one in Australia and one in the US. Because I went and changed my fields of research, completely, which is…very rarely people do that.
Right, right.
And so, they accepted me for what I brought and let me flourish in the field that I chose to work with them…and pretty successful in those areas as well. That is my foundation when it comes to taking teams of diverse people because these people took me from a different field, let me grow, and we have been very successful in that. So in that sense, a lot of mentors come along.
In the business world, there are a few people that have certainly contributed, and most of the work that I talk about is teamwork. Since 1999, from my AOL days to now, it’s not a one-person job because, in the corporate world, teamwork is very important.
Yeah. I was going to say, you underscored it. Clearly, from getting a PhD in quantum chemistry and then even changing it within getting your PhD and then beyond, and becoming a data science leader…you learned from that. You learned that people from diverse backgrounds and diverse expertise can move into the world of data science, if they’re given the right level of mentorship. The fact that you highlighted a bunch of mentors along the way, who happened to be in academia, tells me that that teaching mentality from you has really been a huge part of your career. And I think, for those data science leaders out there, I think that’s a great lesson learned here. You can coach, you can teach to create a high-performing team.
And then, lastly, if I thought that I could get a PhD in only two and a half years, I probably would have gotten my fricking PhD. But that is quite impressive. Most people I know who have PhDs took up to two to three X that.
That wasn’t my PhD. It was not me.
I know, I know, I know. I’m talking about your advisor. Yeah, absolutely.
I worked for three and a half years for my PhD.
That’s impressive, yeah.
I think it’s, again, as I’ve said, shaping of the mindset is very important in this case. And that has worked very well for me.
Well, Dr. Priyadarshy, this has been just a lot of fun. I’ve really enjoyed speaking with you. I hope many of you out there feel the same way. We learned a lot about the teaching background and how that’s influenced your career. We learned a bit about some tips and techniques in terms of how you’re actually presenting those materials and how your background in teaching has helped present materials to folks who are not data scientists, but who are very influential, and people who we need to win over in order for us to keep our jobs.
And then, last but not least, we also learned about some techniques in terms of increasing a diverse workforce, not just diverse in thought and in background, but also from a demographic perspective and getting more women into the data science realm. I think this will absolutely help and create better and more high-performing teams. It’s been a pleasure chatting with you. Thank you so much for coming on. Hopefully, you had as much fun as I did. So thank you very much, Dr. Priyadarshy.
Thank you, Dave. That was wonderful. I if anybody wants to talk about it, sure.
Yeah. Look Dr. Priyadarshy up on his LinkedIn page. I don’t know if you have Twitter, or any other social media, but I’m sure you can be found there. Is there any handles that you want to mention?
Totally the same as my last name, Priyadarshy.
@Priyadarshy on Twitter, send him questions directly. Again, I’m very happy to have you on the show, and thank you all for listening.
Thank you so much.
29:22 | Episode 16 | August 17, 2021
40:04 | Episode 15 | August 10, 2021
38:29 | Episode 14 | August 03, 2021
26:54 | Episode 13 | July 27, 2021
Use another app? Just search for Data Science Leaders to subscribe.
Data Science Leaders is a podcast for data science teams that are pushing the limits of what machine learning models can do at the world’s most impactful companies.
In each episode, host Dave Cole interviews a leader in data science. We’ll discuss how to build and enable data science teams, create scalable processes, collaborate cross-functionality, communicate with business stakeholders, and more.
Our conversations will be full of real stories, breakthrough strategies, and critical insights—all data points to build your own model for enterprise data science success.
Host, Data Science Leaders
