Lekan Adeyeri

I was born in Lagos and moved to Boston at seven. From very early, I wanted to chase my own creative ideas unconstrained, by my technical ability first, and ideally by anything at all. That pull showed up young: on a neighbor’s sporadic wifi, because my parents didn’t want me near the internet, I ran a small YouTube channel that earned about $69 over its lifetime. Coding looked like the way to build whatever I could imagine, so I taught myself, assembling working machines from broken computers along the way. A scholarship to a summer program at Phillips Exeter gave me my first structured exposure to computer science, physics, and discrete math, and from there I started building startups in Boston.

By sophomore year of high school I’d shipped clinical-trial software at eResearch Technology, now Clario: an Alexa and natural-language voice interface that let lupus patients who can’t use touchscreens take part in decentralized trials, still in production years later. Since then I’ve won the UPS National High Tech Award for InjuryEmotion, a computer-vision system for worker safety; placed second of roughly 1,200 international applicants in Twitter’s Early Bird product pitch; co-authored medical-imaging research at Stanford’s Center for Advanced Functional Neuroimaging ( Cornelius G. Dyke Award, ASNR); and worked on a GPT-3 implementation as a software-engineering and product intern at Microsoft. I was considered for, selected for, or a participant in venture fellowships at Bessemer Venture Partners (😅) and Ripple Ventures, and considered for others, including Kleiner Perkins. I once built a grocery delivery platform that reached 450 users at one location on a $30 budget; it made no money 🙁.

At Pitzer College, on a full scholarship, I designed my own combined computer-science and physics major. It won approval from more faculties than the process required, then triggered something the institution had no precedent for: a proposal that normally never goes before a committee instead had to clear two, one of which reached the university president, after which two new rules were written and applied retroactively. I pursued it about as determinedly as a student can. That determination, combined with health issues I did not yet understand in myself, the after-effects of a minor car accident, and an inland-California heat I did not register as extreme, eventually made it sensible to step away for a while. The stretch turned out to be generative. Making sense of my own health sent me to language models to understand the body, and the heat sent me, improbably, toward climate and computational physics. More than that, I wanted room to chase ideas I could feel were about to matter; committing to something I did not actually find interesting felt like the convenient, legible, and wrong choice at a moment when computing was clearly shifting under our feet.

That room is where my current work lives. Under SoarLabs, a self-directed research lab I started, I’ve built seventeen projects in physics-informed neural networks, neural implicit fields, and graph dynamics: surrogates for waves, fluids, cloth, motion, and geometry, all circling one question, when does a learned object become good enough to stand in for the physical process it was trained on. I work the ideas out in writing, too: trained networks as compressed, queryable data structures, persistent geometry versus rendered appearance, and the energy case for neural surrogates. I’ve also begun exploring connectivity infrastructure for places the major networks overlook.

The thread is the same one that started on that neighbor’s wifi: I like making real things exist under constraints, and I’m drawn to problems that are just becoming possible. Every project, paper, award, and archive is sourced here: demo.1001ud.me/references. I’m always glad to talk.

Links

More