I use human-genetics evidence to separate true causal drivers from noise, and I build LLM agents that turn public databases into cited, decision-ready evidence for drug targets.
|
🧬 The science Causal genetics of obesity and metabolic disease Butler-Laporte lab, McGill |
🎯 The method Mendelian randomization + colocalization proteome-wide, to nominate and de-risk targets |
🤖 What I build LLM tool-calling agents over public databases evidence you can check, line by line |
Type a protein and a disease. Get back one evidence card you can check, line by line.
- The card is rendered by code, not written by the model
- The model writes two sentences — a validator rejects the page if either is unsupported
- Benchmarked on 20 pairs history already decided: every GO was a drug that launched
- MR estimates are always retrieved, never computed here
Explore repository → Browse 991 protein dossiers → Card viewer →
Built as the core of the CABS 2026 team project · upstream repo, alongside Natalie Huang's OpenSentinel drug-safety module.
|
McGill University PhD candidate, Quantitative Life Sciences Drug-target discovery · graduating 2027 |
CABS Data Science Summer Intern, 2026 Built a drug-target agent end to end; mentored a teammate to a working one |
Published Nature Communications (accepted, 2026) · Advanced Science (2023) · IEEE TPAMI (2021) ESHG 2026 poster, Gothenburg — proteome-wide causal inference for metabolic liver disease |
Python · R · Bash · Git · Mendelian randomization · colocalization · GWAS / pQTL · LLM agents · RAG · tool calling
Website · Google Scholar · LinkedIn · ResearchGate · Email
Graduating 2027 — open to roles where causal inference and large-scale statistical modelling drive real decisions.