Employee Q&A: Frank Lo, Director of Engineering — Data Science
DraftKings couldn’t produce award-winning products without our Engineering team. We’re excited to spotlight our talented Engineers through our Q&A series. Please welcome Frank, Director of Engineering who’s been with DraftKings for nearly 4 years! Get to know Frank’s story below.
How have you adjusted your work/life balance for ‘shelter in place’?
Well, I have two young boys, Nathan and Nicky. Like others with young kids and childcare closed due to the pandemic, it has been an adventure. Weekdays can sometimes be a juggling act between work and life. I do appreciate seeing my family more often. At the same time, I also miss seeing my DraftKings family in person (not just on Zoom) and am looking forward to the day when we can all safely be back in the office.
How does your team contribute to DraftKings products?
We’re a software engineering division specializing in building algorithm-based products to help with key initiatives on the DraftKings roadmap. Our data science org comprises multiple teams that cover a wide variety of domains and use cases. Some product examples to help illustrate:
- Recommender systems that enable real-time personalization of our Daily Fantasy Sports and Sportsbook apps.
- Marketing decision engines that self-learn how to communicate with users in optimal ways.
- Fraud detection engines that continuously scan activity patterns for suspicious behavior.
- Sports algorithms that predict athlete performance in games, to support DraftKings gameplay operations.
A significant part of our focus has been building our data science tech platform, enabling us to scale our product delivery by allowing multiple business verticals to get leverage from a common core. For example, our Sportsbook recommender algorithms share the same infrastructure backbone as our Daily Fantasy Sports recommenders.
How did you end up at DraftKings? What drew you to the company?
I was actually a super early user of DraftKings, registering my DFS account way back in 2012. I joined the company in 2016. I saw a tech company on the rise, with interesting data science challenges to solve, supporting a fun, engaging product to play. As a big sports fan myself (favorite league — NBA), being a part of a sports entertainment company was something I could get excited about as well. On top of that, the opportunity to start and grow a new data science team scratched my entrepreneurial itch. It has been a fantastic experience so far, almost four years!
You came to DraftKings to establish the Data Science team. What has been the most challenging aspect of establishing a new team?
I had previously started the data science organization at Wayfair in 2011, so this was my second go-around building up a data science group. I see this type of thing as running a startup within a company. Begin with a vision, then hustle to build up demand, gain early traction, and line up new opportunities in a roadmap. At first, the emphasis was on building relationships with internal partners, understanding needs, and being proactive in suggesting new project initiatives that could quickly drive business value. There was intrigue within DraftKings surrounding our new team, but the trust and confidence that we could deliver had to be earned. Thankfully, quick project wins upfront led to meaningful traction in those early days, which enabled us to continue to expand our influence and find creative new ways to drive innovation in the company.
What does a typical day look like for team members?
There are three primary phases we cycle through when tackling Data Science projects.
- Data research — Form an intimate understanding of the data. Study the context behind the problem and decipher the types of narratives hidden away in all the data relationships. Whether beginning a new project or iterating on an existing one, this step is critical towards guiding the design of the solution.
- Algorithm dev — Design and build the computation logic. Methodologies may involve machine learning or other quantitative techniques. Often, simple, but well-architected solutions achieve meaningful results without high complexity cost.
- Software dev — Translate algorithm logic into production-quality code. Prototypes get transformed and refined into robust technical applications that are deployed at scale and have impacts on DraftKing’s end-users.
Our devs don’t necessarily touch all the above in a single day. However, within a project lifecycle, we shift the focus between these different phases to advance projects forward.
What’s your proudest DraftKings moment?
I can’t pick a single moment. I have been most proud when I see our Data Science team members invest wholeheartedly into ambitious projects — and through blood, sweat, and tears, figure out a pathway for their Data Science initiatives to deliver seven-figure impacts on the business. We have had many of these mega wins, but they never come easy. There are always early failures that leave us banging our heads against the wall, wondering why a test algorithm isn’t generating lift over the control. However, there is extraordinary creativity that comes from thinking about the problem all day and night, leading to moments of design epiphany and successful solutions.
What is your favorite language to code in?
Python. It has strong scientific computing capabilities, suitable for machine learning and data exploration. Simultaneously, it is practical as a general-purpose programming language, essential for interoperation within a technical environment and buildout of production-grade applications.
In addition to its versatility, Python is a highly expressive language that is straightforward to write — efficient for developers translating complex ideas into code. It’s also easy to digest from a readability standpoint. One of the best pieces of programming advice ever said to me was “write code for humans (not machines) to consume” (otherwise, you can end up with a repo full of code rot that nobody wants to touch). The Python language upholds this mantra of coding for humans.
The other language used extensively in our data science team is C#. For all the Python love I spelled out, admittedly, it is a comparatively slow language. Compiled languages can run 50x faster. We’ve leaned on C# for production applications where millisecond latency matters.
What do you look for when hiring for your data science team?
When hiring for my team, I want team members passionate about the space, wildly curious, and who never cease to try to think of creative solutions for unsolved problems.
In your resume and during the interview, demonstrate this excitement to us. Here’s a tip on an interview question I often ask: “Tell me about a technical project you worked on *outside* of class or work.” If you were interested and motivated enough about a topic to dive into it without it being ‘assigned’ to you, it shows you have the type of drive to become highly invested in Data Science technical projects and strive to take them to the next level. This intellectual curiosity is the topmost intangible trait I love to see in a candidate, a perfect complement to quant and tech skills.
