Chemical biologist & bioinformatician, MD in progress. I build open-source computational pipelines that ask whether a drug target is actually real — pre-registered hypotheses, SHA-locked verdicts, nulls published, all on free compute. See the full portfolio →

UC Berkeley, Chem Bio · Brandeis, MS Bioinformatics · UAG International MD (MS1)

• falsifiable-targets — content-addressed audit engine for drug-target direction-of-effect (inhibit vs. activate), built on Open Targets colocalization and benchmarked against approved drug–target pairs. 83% acc · 50% cov, failure modes documented not hidden.
• nav17-asymmetry — why the best genetically-validated pain target (Nav1.7 / SCN9A) keeps failing in the clinic: a genetic–pharmacological asymmetry, worked through single-cell, a homeostatic-compensation model, and human genetics.
• esm-trust — when can you trust ESM-C zero-shot variant-effect rankings? A reproducible benchmark showing a model's self-consistency does not predict its accuracy, plus a calibration tool.

• t1d-celltype-of-action — maps type-1-diabetes GWAS loci to the pancreatic cell types they act in, τ-specificity over the HPAP atlas. 21 calls · 145 loci.
• madurella-target-discovery — ChEMBL-driven target pipeline on a neglected fungal pathogen nobody had mapped — and it catches its own artifact: the "top" gene was 384 duplicate HDAC4 records. 10,707 → 7.
• cruzain-in-silico-pipeline — 8-stage open drug-discovery pipeline for Chagas (T. cruzi cruzain): three scoring tracks fused to consensus, selectivity counter-screen vs. human cathepsins, ADMET. Runs on a free Kaggle T4.
• enpp1 — structure-based paralog-selectivity counter-screen for the immuno-oncology target ENPP1 vs. ENPP2 (autotaxin) and ENPP3, ranked by cross-paralog margin.
• degradomap — empirical evaluation of which public-data features actually predict PROTAC E3-ligase tractability.
• loeuf-safety-audit — gnomAD LOEUF-based on-target safety audit: how much genetic constraint argues against knocking a target down.

• ncbi-bioscraper — zero-cost PubMed + OpenAlex + open-access full-text mining in one Colab notebook.
• target-confidence-card — one-page, reproducible target-confidence summary.
• miniprotein_genai — generative miniprotein binder design.
• MCAT-study-app — MCAT concept-practice app for premeds.

• MD — UAG International MD · MS1 (in progress)
• MS, Bioinformatics — Brandeis University · 2025 · GPA 3.85
• BS, Chemical Biology — UC Berkeley · 2020 · GPA 3.67 · CRISPR epigenome-editing thesis

_{pre-registered hypotheses · content-addressed (SHA-locked) verdicts · nulls shipped next to positives · runs on free hardware}