Continuing to pursue a Product Data Scientist career, for me, means also doing a few things that most people in this field don't. From a builder's perspective, building an experimentation platform myself might actually be the builder-shaped version of what that career looks like.
What Antler Was Actually Testing For
I once went through an interview process with Antler about starting a company. The screening step was a 30-second Loom recording; clearing that led to a 30-minute remote interview. The single most memorable moment of that interview was one question — the worst failure I'd actually gone through.
My answer to that question wasn't especially satisfying, even to me. It left me realizing I actually worry about failure more than I would have guessed going in — I think about risk more than I'd assumed. That said, I'm someone who thinks a lot and leans toward research by disposition, but once I've actually decided to execute, I move on it right away. And despite having a Ph.D., there's a builder's disposition underneath it.
Medium risk, medium return, I can live with. High risk, high return — I'm honestly not sure that's actually me, no matter how good the idea sounds on paper.
Antler recruits startup founders specifically, so the interview doubled as a test of whether I'd actually be a good fit as a founder, whether I fully registered that going in or not. Plenty of people go through the same process and come out the other side concluding that a different career suits them better than founder does. I still don't know exactly which career I'll end up in, but it was a genuine chance to think through what disposition a VC-backed startup founder actually needs: super high risk, super high return, swinging for the one home run — invest in ten, and it's fine if nine fail completely, as long as one comes back at 100x or more.
The idea I wrote up and submitted at the time has since turned into something a real startup is actually building — a grading problem, tackled first before anything else. My interest back then was in edtech; these days it's shifted specifically toward experimentation platforms. But the underlying shape of the idea — pick one narrow, well-defined problem and solve that first — is the part that's stayed the same.
Development
I never actually applied to this one, but I did seriously want to learn to build things properly. There are immersive, live-in bootcamp-style programs for exactly that — the kind where finishing one is supposed to turn you into a genuine developer, not just someone who's read about it. What I actually wanted out of it was narrower than "learn to code" in general: I wanted to build a B2B SaaS experimentation platform.
So that's the plan. Build it.