hey,
i'm sure you've been waiting in anticipation for a progress report. this update will depart from the usual format. that's because, i've decided to finally throw in the towel and end the founder journey.
the rationale: simply finding an axis to follow towards a viable business problem-solution proved to be immensely difficult for me. juggling customer discovery, building/prototyping, upskilling as a founder, while trying to grow my domain knowledge made progress absent or imperceptible. i was operating on 6-8 week ideation/validation cycles but this was simply to short of a period to gather meaningful market signal for whether something would work or not. finally, i had to ask myself, is there any idea I believe in enough to dedicate a minimum of 6 months to deeply explore? the answer was no.
looking back: after the first startup blew-up on the tarmac, and after passing through the gauntlet of co-founder dating, deciding to forge ahead solo (mistake #1), in the chosen domain (mistake #2), was fatefully wrong. i really was not setup for success. i had a search space (too broadly defined as "ml tooling/infrastructure"), but no direction. i had some domain experience, which was decaying in a problem/design space that was rapidly shifting. i had pedigree, but a weakly constructed support infrastructure (in the form semi-routine checkin with peers and informal advisors). i had the technical chops, but not the resources to match the ambition of many of my pursuits. all combined, i was navigating too wide of a search space while moving too slow. the plan was to eventually team up with a co-founder once a customer/problem had been set, but i was unable to get to that point on a practical timeline.
up next: i made the decision to move on at the beginning of the year, and for the following act, i'm hunting for jobs in the space of ml infrastructure—which i'm still bullish about. my search has been strategic, shaped by the following criteria: (1) work at a category leader/cutting-edge company, (2) where I can build on my pre-existing ml expertise, (3) that has sufficient resources/leverage for high impact, (4) while fostering a talent-rich, fast-pace, high-agency environment. yes, this includes many of the AI labs, but I'm also really interested in the more obscure startups/organization pursuing unconventional, but wildly interesting ideas. if you have line-of-sight towards one of these gems, please lmk :).
wrapping up: from building an AI agent for immigration, through helping startups with large-scale pretraining runs, shipping an LLM tool-use harness, fine-tuning small models for startups, attempting to build an ML compiler, and more, there were plenty of moments of growth and fun. but what i'll treasure the most is the people i met along the way.
for anybody that wants to learn more, i might follow-up (no promises) with a more thorough post-mortem. but also, you can also reach out to me with your questions and curiosities. as for this mailing list, i'm going to merge it into my newsletter followers (lmk if you want to unsubscribe). this is not goodbye, it's see you later ;)
thanks for tagging along for the journey — onward and upward!
hardik