Staff-level ML Engineer with 12 years of experience across Knowledge Graphs, Proteomics, and Generative Biology.

I build end-to-end discovery platforms for high-dimensional biological data. My focus is on architecting the end-to-end framework —from initial hypothesis modeling to production-grade discovery engines—ensuring that high-scale biological insights remain both reproducible and biologically consistent.

• GenBio & scRNA-seq: Scaling and benchmarking Foundation Models (scGPT, Geneformer) for cell-type annotation and gene perturbation.
• Agentic Systems: Designing stateful multi-agent RAG and dialectic reasoning loops via LangGraph (see Cherchoux).
• Proteomics & MTL: Architecting Multi-Task Learning frameworks for population-scale proteomics (UKB-PPP), optimizing for rare-phenotype signal retention.
• Relational ML & Graphs: Large-scale Knowledge Graph modeling (Neo4j/CERN) and Link Prediction for therapeutic discovery.
• Cloud & MLOps: High-autonomy contributor in GCP/AWS environments with a strict focus on reproducibility and data integrity from ingestion to inference.

• scFM Benchmarking Library: Architected a unified evaluation framework to standardize the fine-tuning and embedding generation of single-cell FMs, reducing R&D experiment cycles from weeks to days.
• Validated Link Prediction: Developed a relational ML pipeline for drug-target discovery. Results were selected for wet-lab validation, identifying novel therapeutic candidates.
• UKB-PPP Multi-Task Engine: Scaled a proteomics embedding pipeline to handle high-dimensional, multi-phenotype population data.

• Contributor: edx-platform (Python/Django) – contributed to infrastructure serving millions of learners globally.
• CERN Alumnus: Designed large-scale bibliometric and citation graphs using Neo4j.
• Remote-First Engineering: Over a decade of experience in distributed teams. I specialize in co-designing automated discovery workflows that enable rapid scientific iteration while maintaining the rigorous data integrity required for therapeutic discovery.