Bioinformatics scientist — 8 years applying ML and software engineering to biology, 6+ in oncology drug discovery at Arcus Biosciences (tumor-immune phenotyping, neoantigen & TCR workflows, multi-omics platforms and pipelines).
I build biology agents and the evaluations that test whether they run real analyses with calibrated uncertainty — systems that recognize when the evidence isn't there and stop, instead of returning a confident answer that's quietly wrong.
Selected work
- coscientist — multi-agent hypothesis tournament with a truth-tracking eval harness grounded against GPQA-bio
- ai-scientists — trust profile of biology agents on BiomniBench-DA (contamination, calibration, refusal, benchmark validity)
- methods-graph — bioinformatics methods knowledge graph (EDAM + nf-core + BioContainers on Kùzu)
- dogma — graph-grounded bioinformatics method guidance in the IDE