Software engineer. Applied AI, healthcare, and scientific systems.
| Project | Focus |
|---|---|
| clinicalRAG | Grounded clinical Q&A: LangGraph pipeline, per-sentence citation attribution, grounding score, refusal guards, offline evals |
| agent-reliability | Tool-using agent harness: JSONL traces, deterministic graders, and a CI gate that fails on failure-class drift |
| vesicle-seg | 3-D EM segmentation: Zarr storage, overlapping-chunk inference with inner-region writes, residual U-Net against a threshold baseline |
| motioncode | Waveform classification: record-grouped splits with leakage tests, feature baselines at 0.733 accuracy against a 0.467 CNN |
| matchcare | Multi-tenant FastAPI/React SaaS: bcrypt/JWT auth, role dependencies, org_id scoping enforced in the queries |
| PLAN-IT | Flask API and Expo client for plant recommendation, built with a team at UT Austin and later extended |
syncura-readmission (readmission risk with SHAP explanations) · trinity-care (senior-care Expo + FastAPI app) · merchantlens (market-basket lift analytics).
Python · FastAPI · TypeScript / React · Expo · LangGraph · PyTorch · pytest · Docker · GitHub Actions