Selected Projects
A timeline of platforms, models, simulations, and bench-to-computation projects from the plan. The through-line is scientific data work that turns into usable systems.
Multi-agent research pipeline that proposes a mechanism-of-action hypothesis for a candidate scoring feature, dispatches an adversarial pass to try to disprove it, and grades what survives across five evidence axes before it reaches production scoring.
Built and own the central inference platform delivering every microbiome health score to Viome customers and clinicians, turning internal scientific packages into revenue-critical production infrastructure.
Fully local macOS meeting recorder, scaffolded and architected by hand with AI writing the bulk of the implementation.
Rewrote O(n2) parsing logic to O(n), cutting ODM Parquet sync from 3 days to 8 hours and shrinking a core operational bottleneck.
Built the internal developer platform that standardizes scoring release workflows behind compiled Python CLIs, reducing manual release effort and making deploy, rollback, notify, export, and ID translation repeatable.
Authored the internal Python packages for microbiome score development and built explainability tools that reduced multi-database score audits from 30+ minutes to self-service lookup.
Led AbbVie-facing drug-discovery platform work, advanced 5+ leads to CRO validation, and rebuilt HR-Algorithm as a 1000x faster Python package integrated into Molsoft ICM.