CODE Corner: Profiling HBS students and their side projects every month. This month: Bansi Shah (HBS ’17, Section A) shares on Robots, Artificial Intelligence, and Product Management Internships for the summer.
Anshul Bhagi: Why did you study Computer Science?
Bansi Shah: I always wanted to be a power ranger, but given the limited job prospects there, I found the next best way to work with amazing machines that looked like Zordon’s energy tube.
AB: We hear you’re into robots and artificial intelligence. Why? How did that happen?
BS: In 2008 (as an undergrad) I interned on a high-frequency trading desk. I helped write and test black box algorithms that traded securities on public markets. We measured performance by seeing how much better we did than our human counterparts at the firm. For most of the summer we killed it, but then the 2008 credit crunch hit and we lost most of the money we made that summer. The algorithms were trained on datasets representing a tiny slice of history. Humans were better at using intuition to decide how and when to trade in uncharted territory. I became pretty excited about considering the ways computers could massively outperform humans and, in contrast, the ways humans still had an edge.
AB: Tech evolves rapidly — What do you read to stay on top of it?
BS: I like to read a bit of mainstream tech news (TechCrunch, HackerNews) just to hear what is new. But beyond that, I like to read a lot of science fiction (check out Neal Stephenson), trend forecast reports (check out Mary Meeker and Ping Li), and essays from tech thought leaders (check out Peter Norvig and Paul Graham).
AB: What is a PM? How does one get a PM gig for the summer?
BS: There are tons of great articles on Medium, Quora, and company blogs that provide great (but highly varied) definitions of a product manager. The best one I’ve heard is that a product manager is the singular, unbiased advocate of a product. Everyone else in a company views the product from a bias. A salesperson wants it to have a feature that helps close the next deal (“Can we add single sign-on to this?”), an engineer might want to just build the next cool thing (“Can we build a model to predict where the user will click next?”), and a marketing person might want it to align with the next flashy trend (“Can we use big data?”). A product manager spends a lot of time listening to everyone in the company, plus the competition, the sector, and especially the users. Then he/she makes a decision about what should be built next.
AB: What makes a good PM?
BS: A PM should have an amazing sense of the simplest set of features that provide the desired experience. “Franken-products” that have a laundry list of features tacked on (Excel export, Twitter integration, rich text formatting, or collaboration tools) are overwhelming. They create code complexity that grows exponentially into an untamed monster code base. That overhead slows the entire engineering team down. This happens because every time you add a feature, you have to think not only about that feature but also about how that feature interacts with every other part of the code base. Good PMs recognize that a product is great when you cannot remove any more features without ruining it.
AB: It’s hard for non-technical people to get PM gigs, what advice do you have for first-time PMs as they go through recruiting for the summer?
BS: Some of the greatest PMs I’ve met are not necessarily technical. They just know everything there is to know about the relevant industry, the product, and the users. A PM should bring something to the table that the engineers are not bringing. He/she should not be spending time to develop a really deep understanding of the product’s guts and inner workings, because that would be redundant. Instead, they should provide value by understanding how, where, and why that product interacts with the rest of the world. Ultimately, I believe the real reason why companies like technical PMs is because they don’t want PMs to get bullied by engineers. A few short term tips are:
- In interviews, emphasize past experiences that have given you a detailed, unique perspective on the industry or user for the product you may be working on
- Each week, write a critique for a product. For each product you choose, form opinions on features you like, dislike, and what you would change and why.
- Try building something from scratch, even if it’s simple or just built on top of drag-and-drop platforms or sites such as Weebly / Squarespace / Shopify / WordPress. Think carefully and purposefully about the experience (why put the menu bar in the top left? What is the first thing a user should see when opening the app?)
AB: What are you working on right now?
BS: An algorithm that uses natural language processing to automatically generate cliff notes for BGIE cases…just kidding. Last semester I helped write an algorithm that suggested phrases to autocomplete notes for medical records. For example, if you are writing discharge notes, the probability that “ventricular” comes after the word “left” is over 70%. In our (pretty unscientific) empirical alpha test, we sped up note-taking by 20%.
AB: How are you spending the summer? What’s next for you?
BS: I’m helping a VC firm evaluate software and hardware opportunities in the self-driving vehicle space. Hopefully, this technology space won’t see the same downside as algorithmic trading.
Established in 1937, The Harbus News Corporation is the independent student news publisher of Harvard Business School.