David Yang
CS student at UPenn focused on building actually useful tools. Been gripped by the power generation space as of late. Building now, launching soon.
Highlights: Founding Engineer at Idler (YC S'25) shipping coding RL envs to frontier lab; previously published at ICLR with Reticular AI (YC F'24); interned at Amazon with an agentic project that was presented to VPs.
On my blog I've written about low-level systems fundamentals, past work experiences, and interesting projects.
Experience
Publication
Towards Interpretable Protein Structure Prediction with Sparse Autoencoders
ICLR Workshop · Mar 2025
Scaled SAE to ESM-2 3B (ESMFold backbone), enabling interpretability of structure prediction for the first time. Revealed that structural determinants are very sparse — 8–32 active latents/position needed for reconstruction.