On this episode of The Artists of Data Science, we get a chance to hear from Brandon Quach, a data scientist who has a PhD in bioengineering, and has worked on threat analysis for security and business ecosystems. He's currently a principal data scientist and manager, leading the charge to modernize the customer experience by applying machine learning to customer support.
Brandon shares his perspective on how data scientists should approach problems, the importance of passing on knowledge, how to be a leader in the data workspace, and the appropriate mindset to develop when faced with difficult problems. Speaking with him was an honor, and this episode has something for everyone to take away from.
Some notable segments from the show
[11:50] Brandon discusses automation and whether or not we will be able to automate human judgement
[18:01] What qualities do you need to become an intrapreneur in your organization
[22:19] A unique way to approach leadership in your organization
[30:08] Why great thinkers abhor being told what to do
[37:37] How important is agile and scrum methodology in data science
[46:13] The mindset you need to accept the monsters in your life
Brandon’s journey into data science
Brandon’s journey began during his PhD program at Caltech, where he was focused on experimental subjects. While working on these experiments, he began to notice factors in an experiment that had large impacts, but were not quite being measured. He thought about the math behind these factors, which led him to becoming a consultant. In this role, his title was Senior Associate, even though he was working as a data scientist. During this time, the field of data science had not quite been established, hence his title.
[4:06] “Yeah. I mean, for me, there wasn't really much of a moment of breaking into it because there wasn't too much of a field when I started. Right. So I did my PhD at Caltech and I was very experimental. We were taking silicon wafers and we were trying to make them do things like duplicate DNA, using micro-fabrication, things like this throughout the process. I was thinking there were a lot of setbacks that happened in the laboratory setting that wasn't didn't really reflect what I thought I was capable of intellectually. Right. So, you know, maybe I would design a cool experiment, but by the time I did it, then something random would happen.
Maybe I turned on the nitrogen tank. There's no more nitrogen, and everything has been thawed, everything's been prepped, and now it's like you've just lost a whole day's worth of work. And sometimes worse, sometimes weeks or maybe even months worth of work because of that one moment when everything ready to go. Something happened that totally contaminated the experiment. And what things like that happen - and I kind of noticed well, you know, I think I'm more of a like a thinking person. I think I enjoy doing the math more than the experimental stuff.
So I started to look into sort of alternate careers that weren't so experimental. And, you know, consulting was there. There were the legal path was there. There were legal firms would come in and say, well, maybe if you work with us reviewing patents and such for a few years, then you might go to law school after that and and become like a patent lawyer. And so then I was interested in all those all of those things. And eventually I went into consulting. So it was with my first employer Opera solutions now known as ElectrifAi. And we just did consulting for a bunch of different companies. There are a lot of different fields that you had mentioned before, and the title was called Senior Associate. That was just it. And it wasn't really called - I mean, even when I left...it was still called something like analytics manager. It wasn't a data scientist. We never really called it that until probably that transition of when I left and went to Lytx when the title became data scientist.”
Where is the field headed in 2-5 years?
When Brandon entered the field, everyone was saying that automation was going to allow models to build themselves, and he thinks we are still in this phase. He thinks this is going to continue to happen over the next few years as well.
[12:06] “Yeah, you know, I thought about that a lot. I'm not really sure myself, because two to five years ago, everyone was saying that this automation was going to come in, that this building the models, the models were gonna build themselves, they're going to tune themselves and all this.
And it makes sense to me. A lot of the things that I was doing did seem pretty simple. You would do it a couple of models you would choose, you would optimize in this way. You would choose this. You would do this parameter searches and. Yeah. You could. I mean, I was automating them to write. I was writing scripts that would automate all those kinds of stuff. And I thought, yeah, this is probably got legs and this is gonna happen. And so for the last two to three years, I was thinking the next to two years was gonna be about automation and that the data scientists would be akin to a modern, let's say, mechanical engineering who might have done studies in how to - like in fluid dynamics, right. And how to model fluids and what's the pressure and velocity at every point along this wing. But they have software for that. You do do the simulation and you're like, well, now I've got the software. So you're thinking, does that mean I don't need the engineer because the software did it automatically? We're - I think we're gonna get to that phase. But the strange thing to me is that I've gotten the impression that that phase is coming very quickly.
For the last couple of years. So now here I am, right. Fast forward two years and it's not and it's kind of here. I've seen it here and there, but I'm at least I'm still using Python. I'm still coding things myself. And so I think that's what's going to happen, continue to happen in the next few years, two to five years. That being said, you know, maybe the time will come when a lot of these efforts to automate things come into play. And, you know, as I mentioned before, it depends on the industry and the problem you're working on in my career path, since I've always been working on new problems. I don't see it impacting us much, but I'm trying it right. And some of the employers that I'm working for, they themselves are trying to build and have built these kinds of things. And I'm I'm trying to use them as well. And I'm providing feedback on the features that have been developed. And what does it mean to work on a like a real data science project where our ROIs expected it to happen and not just kind of a research thing?”
What will separate great data scientists from the rest of them?
[15:56] I think it's the ability to think through problems, and having the intuition to think about what the next step is to your problem.
Key takeaways from the episode
Important soft skills
[50:24] Learn how to think through a problem in its individual components.
How to be an intrapreneur
[18:16] An intrapreneur is somebody who is willing to do whatever it takes to solve the problem, even if that means thinking like a software engineer or from a business perspective. You need a vision, and the ability to take ownership.
Growth mindset and grit
[44:17] You show grit in all aspects of your life when you learn to live with setbacks, pain, and the work that’s required in any endeavor.
[22:30] My philosophy on leadership is based on servant leadership. This is a leader that helps people grow and produce their best work, and gives people the independence to choose how they solve a problem, rather than telling them what to do.
Mindset for difficulties
[46:13] If you have monsters and you feed them, they will keep coming back. Even if you don’t feed them, they may not go away. Don't think “woe is me” for the problems in your life. This feeds the monster. Instead, just accept them and move on.
[22:37] “...trust, to me, comes from your ability to not be scared of the results that come out of your work or anything that you do.”
[27:25] …”If I received good advice and….good guidance, then I feel it's sort of my job, my duty, to pass it on to the next generation”
[30:08] “Great thinkers like to figure things out and come to a point that they believe in the solution.”
[35:33] “I want people to look back long after I've gone and say...that decision that was made early on that nobody had appreciated...turned out to be really critical down the road…”
[53:33] “...successful data scientists can think through any kind of problem surrounding data science, not just the core problem.”
[57:05] “You should learn how to think through code. How can you learn how to think through code?. Well, either you have a built in imagination... and/or you have gone through a lot of iterations of code and you can understand the process...”
The one thing that Brandon wants you to learn from his story
[58:21] Expect that good and bad things will happen to you. The good and the bad things will stick with you through your journey, and you can’t get rid of them. Just accept that, don’t fight it.
From the lightning round
When you need to communicate a problem with your leaders, make sure you bring solutions.
Source of motivatio:
What motivates me is the idea that I'm going to do something that I'm proud of, not something that I hope somebody else will like. As a leader, what motivates me is to help people grow.
Advice Brandon would give to his 20 year-old self
Keep doing things that you think are interesting.
Topic outside of data science we should study:
Body language. Data scientists spend time explaining things to non-data scientists. If you can understand body language, you may pick up on cues of whether people understand you or not.
Books that Brandon recommendeds you to check out
“Case in Point: Complete Case Interview Preparation” by Marc Cosentino
Books and other media mentioned in this episode
“Search Inside Yourself” by Chade-Meng Tan
“Linchpin: Are You Indispensable?” by Seth Godin
How you can connect with Brandon Quach
Connect with Brandon on LinkedIn
Personal Website and Blog