MS Data Science — Stevens Institute of Technology — graduating May 2026
I build ML systems that go to production, not notebooks that collect dust.
Currently: Claude Code, Agent Orchestration, RAG pipelines, LLM agents, trying to keep up with the latest tech...
Email ・College Email ・ Linkedin ・ Portfolio
| Problem | Result | Stack | |
|---|---|---|---|
| Insurance Policy Q&A Agent | Insurance docs are dense — people can't find answers fast | Two-tier RAG: local vector search for fast answers, auto-escalation to Gemini for complex cross-section analysis. safety guardrails catch sensitive topics before LLM responds | LangChain ChromaDB Ollama/Mistral Gemini Streamlit |
| Lumbar Spine MRI Classification | Radiologists manually review hundreds of MRI slices per patient — slow, subjective | Benchmarked 5 architectures. EfficientNet-B3: 75% accuracy, 57% severe-class recall (2× next best) with fewest params among transfer models | PyTorch timm albumentations pydicom DICOM→2.5D |
| Exam Score Predictor (MLOps) | Demonstrate full ML lifecycle, not just a notebook | EDA → 8 models → Ridge (R²=0.88) → Flask API → Docker → GitHub Actions CI/CD → deployed on AWS EC2 via ECR | scikit-learn Flask Docker GitHub Actions AWS |
| Time Series Forecasting | Compare SARIMA vs automated methods on real data, analyze COVID recovery asymmetry | Box-Jenkins models beat automated ETS on both datasets. financial markets recovered from COVID in ~6mo, consumer behavior took ~24mo | R forecast Shiny plotly — live dashboard |
more projects in pinned repos below. new work added regularly.
also: LangChain · Claude Code · Cursor · ChromaDB · FAISS · Hugging Face · OpenAI API · Gemini API · XGBoost · MLflow · Pandas · NumPy · Streamlit · Tableau · Airflow
looking for data scientist / ml engineer / ai engineer roles — open to full-time starting mid-2026. let's talk.