Bio
Peter is a machine learning researcher and bioinformatician focused on AIxBio for biosecurity. As a Senior Research Scientist at SecureBio, he builds safeguards and evaluations for frontier AI models to help prevent future pandemics arising from AI misuse. His current work focuses on building benchmarks for "superhuman" AI capabilities, assessing how well AI models can predict outcomes of biological experiments.
Previously, during his PhD in computational biology, Peter developed multimodal deep learning models for biomedical data, including for cancer diagnostics.
Projects
Research Direction: AIxBio for Biosecurity
Mentored projects
- An agent that refreshes biology-experiment evals from recent publicationsQ3 2026, AI Safety FellowshipOngoing
- Automated adversarial red-teaming of frontier model CBRN safeguards, with particular focus on dual-use biological researchQ1 2026, AI Safety Fellowship
- Testing reproducibility of narrow biological AI tools with coding agentsQ1 2026, AI Safety Fellowship