DeepChem
Open-source deep learning platform for drug discovery, materials science, and biology — research architect and GSoC mentor.
Most researchers who want to apply deep learning to molecular problems hit the same wall first. Before they can ask a single scientific question, they have to rebuild the infrastructure to ask it: data loaders, featurizers, model plumbing, reinvented in lab after lab and rarely shared. DeepChem exists to knock that wall down. It's an open-source deep learning platform for drug discovery, materials science, quantum chemistry, and biology, built so that anyone, anywhere, can reach for the tools instead of rebuilding them.
I've served the project as a research architect and Google Summer of Code mentor. GSoC pairs open-source projects with student contributors for a summer of real, shipping work, not toy problems. Over multiple cycles I've worked alongside ten-plus PhD students and postdocs on the platform's core capabilities, from neural ODEs to conformal prediction to transformer models for genomics. Together they've contributed thousands of lines of code to an ecosystem the whole computational chemistry community draws on.
The mentoring taught me more than the code did. Early on I was too prescriptive. I'd hand a mentee the whole plan and watch them execute it competently and learn almost nothing. Hand someone a map and they follow it; hand them a compass and they learn to navigate. So I started setting the problem and shaping the environment, then getting out of the way. The best moment in any cycle is when a contributor stops executing my version of the project and starts executing their own.
DeepChem matters because it proves a model: open infrastructure, shared freely, produces better science than proprietary tools locked behind licenses. The code is open. The mentorship is open. The knowledge compounds, person by person, pull request by pull request. That's the same pattern Lattice Protocol extends to federated research at scale, and the same conviction Bharath Ramsundar built DeepChem on in the first place. If you have been waiting for permission to work on open-source science, consider this it. Come read some code with us.