Bioinformatics × Clinical AI × Agentic Systems
M.S. Bioinformatics @ Johns Hopkins · B.A. Bioinformatics @ UC San Diego
Building AI that actually ships in clinical settings.
-
CTO @ Continuum Health — Full-stack clinical documentation platform replacing legacy care-plan workflows for an Adult Day Health Care center. Automated 116-page IPC reassessment packets (40+ hrs → <5 min), Azure OpenAI dictation + SOAP note generation, full billing pipeline across Kaiser / Medi-Cal / LA Care.
-
CTO & @ C-URE-PAIN — Clinical AI platform for pain diagnostics: real-time 3D WebGL pain mapping, XGBoost/RF ML pipeline on 1,500+ physician-labeled cases (78% balanced accuracy), LangGraph agentic orchestration with human-in-the-loop review, automated CPT/ICD clinical reports.
-
Agentic AI Scientist @ A*STAR IMCB (Singapore) — LLM-agentic workflows over Cox survival models trained on the Singapore Longitudinal Aging Study (~6k participants). RAG-grounded chat, lifestyle simulation, biological age platform deployed at international conferences.
ML / AI
Languages & Frameworks
Infrastructure
| Project | What it is | Stack |
|---|---|---|
| BioAge Platform | Biological age prediction from 5 validated models (Cox survival, functional, body systems, frailty, cognitive). Trained on SLAS 2,778-participant longitudinal study. Deployed at A*STAR conferences. | FastAPI · React · OpenAI · RAG · AWS |
| Continuum Health (private) | Clinical documentation platform for ADHC center — IPC care plans, AI-generated SOAP notes, billing pipeline | React · TypeScript · FastAPI · Azure OpenAI |
| C-URE-PAIN (private) | Pain diagnostic AI — 3D body mapping, XGBoost triage, LangGraph agentic orchestration, automated clinical reports | React · Three.js · XGBoost · LangGraph · FastAPI |
| HLA Compatibility ML (Immunomatics) | Biologically-informed ML pipeline for 30K+ HLA allele pairs — BLOSUM62 features, Siamese embeddings, 78% AUC-ROC | Python · scikit-learn · pandas |
| GRPO Policy Optimization (Johns Hopkins) | Multi-turn RL system for LLM sequential decision-making; GRPO-style optimization via Hugging Face TRL + Unsloth; 40% improvement over supervised baseline | PyTorch · HuggingFace TRL · Unsloth |
| BC-Design Benchmarking (Gerstein Lab, Yale) | 6-metric evaluation suite for inverse protein folding; curated 5K+ PDB/CATH structures; diagnosed 40% degradation on sequences >500 residues | PyTorch-Lightning · BioPython |
I'm a bioinformatician who ended up deep in clinical AI engineering — building systems that go from raw data to production decisions in real healthcare settings. My work sits at the intersection of:
- Agentic AI — LangGraph orchestration, RAG, tool-use, RLHF/GRPO
- Clinical ML — survival models, diagnostic triage, HIPAA-compliant pipelines
- Bioinformatics — protein structure, HLA immunogenetics, longitudinal aging studies
Currently finishing my M.S. in Bioinformatics at Johns Hopkins while building at two healthcare AI startups.