All case studiesAI Evaluation
Adversarial AI Evaluation — Scientific Reasoning in Bioinformatics
Mercor · Independent contract · Frontier LLM benchmarks
At a glance
- Client
- Mercor · Independent contract · Frontier LLM benchmarks
- Area
- AI Evaluation
- Scope
- LLM Red-Teaming, Failure-Mode Analysis, Synthetic Clinical Datasets, NGS Formats
Result
Frontier-LLM limits mapped in genomics, transcriptomics and epigenomics reasoning, with mitigation strategies
What I did
Adversarial evaluation of frontier LLMs on scientific reasoning in bioinformatics (genomics, transcriptomics and epigenomics). I design experimental tasks and synthetic clinical datasets, probe failure modes, build reproducible multi-agent harnesses and execution environments, and validate and document the results to find where models break and how to push past those limits.
Technologies and topics
LLM Red-TeamingFailure-Mode AnalysisSynthetic Clinical DatasetsNGS FormatsMulti-Agent HarnessPython