CoScientist
Scientific hypothesis generation
A multi-agent research system, with evidence checking and an experimental biotech-company structure.
Research prototype
Can collaboration between agents produce better scientific hypotheses?
Biology · agents · evidence
Computational biology
and scientific AI
I’m a bioinformatics scientist. I build computational methods and research systems for biological discovery.
I work on biology agents, benchmark design, and checks for scientific methods and evidence.
Projects
Research programs
ai-scientists · CompBioBench · Benchmake
Biology-agent evaluation
Auditing tasks, scoring reasoning, and building evaluations that expose failures in scientific work.
Evaluation in progress
How reliably do benchmarks measure biology-agent performance?
Dogma · methods-graph
Methods & evidence controls
Connecting causal hypotheses and evidence to explicit method assumptions, diagnostics, and data handoffs.
Local integration
Can method checks identify invalid analyses before execution?
Notebook
Recent notes
Experiment results, failed approaches, and next steps.
Auditing eleven unanswerable task variants
Six initial variants needed correction before their refusal results could be interpreted.
Task validity, calibration, and judge disagreement
Historical BiomniBench-DA experiments on contamination, cohort design, workflow consistency, refusal, and scoring.
Elo and correctness on GPQA-bio
Sample size, aggregation, and uncertainty in the historical Elo–accuracy analysis.
Experience
Bioinformatics and
oncology research
My experience includes chemoinformatics, behavioral genomics, and bioinformatics across discovery, translational, and clinical oncology.
Background & experienceContact
Roles and collaborations
I’m interested in techbio AI roles and collaborations in biological research, research engineering, and agent evaluation.