Applied research
Built a RAG pipeline over the biological foundation model literature to systematically map emerging use cases, adjacent tools, and workflow patterns across modalities. The analysis fed application discovery and platform strategy.
At Biohub (formerly Chan Zuckerberg Initiative), I work across product, AI research, and cell science, partnering with computational biologists, ML researchers, and engineers to turn emerging research capabilities into products and scientific workflows.
Releasing biological foundation models
Coordinated the release of nine biological foundation models to the Virtual Cells Platform, across internal launches and external submissions. I worked with AI Research, Engineering, Product, and Communications on model evaluation, technical documentation, and release preparation. The work grew from integrating external models to releasing models from our own research teams.
Releasing a model meant working through what a researcher would need to use it: a clear description, a working example, documented inputs and outputs, and a way to report problems. I worked with model developers and the platform team on those requirements, contributed to submission and QA standards, and developed technical tutorials. Developer feedback helped us identify gaps in the platform as well as in each release.
I partnered with AI research and engineering teams during biological foundation model development to evaluate emerging capabilities, identify applications, and translate technical progress into product use cases, demos, and roadmap recommendations.
Additional Work
Biosafety and evaluation
Developed biological AI safety analyses presented by CZI leadership to the Frontier Model Forum, including work on protein language model safeguards. I also evaluated a proposed biosecurity control for protein search, measuring its effects on legitimate research workflows to inform the release decision.