Building the Virtual Heart.

AI/ML engineer with bits of bioinformatics. Europe.

I build domain-specific AI/ML for the life sciences — where the biology decides the architecture, not the other way around. Constant learner, challenge enthusiast.

Co-Founder & Head of Research at Applied Scientific Intelligence (ASI) — a cardiac-first drug development lab. ASI pairs prediction models with mechanistic simulation to build a Virtual Heart that catches cardiac toxicity before it kills a program — and redesigns the compounds it already killed into new, safer drugs.

1. Alexandria — #1 scientific-literature agent on LitQA3, FigQA2 & TableQA2, built with NVIDIA. Best AI Scientist award @ ICML 2026.
2. CardioSafe — multi-channel cardiac safety model (hERG, Naᵥ1.5, Caᵥ1.2, IKs); AUC 0.917, beating published baselines.
3. Virtual Heart — CardioSafe's predicted currents driven through cardiomyocyte simulation: Torsades risk scored in silico at 96% specificity, no assay required.
4. RefusalBench — open benchmark auditing how often frontier models refuse legitimate biological research.
5. VCBench — multi-dimensional benchmark for evaluating single-cell foundation models.

More on ASI