I build AI systems that ship. Currently building biomedical NLP pipelines for a genomics platform at Sivotec Inc. and doing applied AI research at FAU.
My work spans clinical ML, agentic AI, and biomedical NLP. Before that, I spent two years building Snapchat AR lenses. 300+ lenses, 219B+ views. That taught me more about shipping fast than anything else.
Open to full-time AI/ML Engineering and Data Science roles starting early 2027.
| Project | Description | Stack |
|---|---|---|
| MOSAIC | Fairness-aware CVD risk prediction on 551K+ NIH patients. AUROC 0.93, 74% reduction in Black-White AUC gap. FastAPI clinical decision support with SHAP explainability and a 4-agent Gemini pipeline. FAU All of Us Hackathon Winner. | XGBoost, SHAP, Gemini, FastAPI, Next.js, BigQuery |
| DARWIN-PHOENIX | Co-evolutionary LLM framework where a generator and adversary compete to produce better code. Introduced behavioral fingerprinting via TF-IDF cosine drift as a degradation signal. 656 runs, 4 conditions. arXiv preprint. | LangGraph, Qwen3-32B, HumanEval+, Python, Docker |
| CareGuide | Deterministic 5-agent pipeline that de-identifies EHR records, grounds care plans via RAG on USPSTF guidelines, and generates a gamified Health Engagement Score. | LangGraph, RAG, ChromaDB, Gemini, Streamlit |
| OURI Assistant | Gemini-powered Next.js app connecting FAU undergraduates with faculty research opportunities. Built for and deployed at FAU. | Gemini, Next.js, Vercel |
| Stochastic Cliff Walking | Benchmarks SARSA, Q-Learning, and Double Q-Learning under stochastic wind and slip dynamics with theoretical ground truth via Model-Based Value Iteration. | Python, OpenAI Gym |
Python PyTorch TensorFlow Keras Scikit-learn XGBoost HuggingFace SAPBERT LoRA/PEFT
LangChain LangGraph DSPy RAG ChromaDB Gemini API OpenAI API
FastAPI Docker GCP Next.js React Streamlit SQL Power BI