From translational cancer immunology to biologics manufacturing, I build systems that connect data, decisions, and teams.
GNN models (GCN, GAT, and message-passing variants) on CosMx LBCL datasets to forecast CAR T response weeks before infusion.
Outcome: Surfaced spatial immune niches linked to relapse; guiding patient stratification experiments.
Built an end-to-end candidate triage system: sequence cleaning, structure-aware embeddings, and ranking models for affinity prediction.
Outcome: Improved hit rate and reduced experiment turnaround by 30%.
Designed adaptive control algorithms and telemetry dashboards for large-scale biologics manufacturing.
Outcome: Stabilized runs and cut manual overrides during pilot batches.
Mapped resistant tumor microenvironments across LBCL and DMG using multi-omic single-cell and spatial assays.
Outcome: Identified chemokine-driven myeloid states that drive CAR T resistance.
Signal processing and visualization suite for gastrointestinal motility recordings.
Outcome: Adopted by the motility clinic for prospective studies.