# The model is the product here. LLM apps, RAG pipelines, computer vision, and applied ML, each shipped end to end.
AI-powered triage system for healthcare, winner of 1st place at Rutgers HealthHack 2025. Currently in development with Rutgers through a completer program to bring it to production.
Voice-first AI learning platform that turns uploaded material or any topic into a personalized roadmap with guided lessons, practice, assessments, and progress tracking.
Local AI assistant for Obsidian that answers from your notes with BM25 search. Runs on localhost with multi-tab sessions and swap-in backends like Claude Code or Codex.
LLM-driven platform that generates quizzes and tests from user-provided material, winner of 2nd best overall and best educational hack at HackUMBC 2024.
Pantry tracker that flags food before it expires, then suggests recipes and donations for whatever needs using up.
Capstone system that develops and evaluates persona-consistent LLMs grounded in Hamlet and Macbeth via LoRA fine-tuning, RAG, and a FastAPI + React demo for single-character and multi-model dialogues.
Student analytics dashboard built on a Neo4j knowledge graph, looking for patterns in academic habits and engagement.
Decision-support dashboard for a poultry farm client that analyzes litter survey data to highlight what drives performance and flag farms at risk of underperformance.
Camera-based counter for newly hatched chicks on a conveyor belt, built to handle the overlapping birds that throw off a laser counter. Patent consideration.
# Prefer grounded retrieval and editable outputs over black-box chat. Keep humans in the loop for high-stakes domains (healthcare, education, farm decisions), and favor local or inspectable pipelines when the data should not leave the machine.
Python
PyTorch
TensorFlow
HuggingFace
LangChain
OpenAI API
Google Gemini
Scikit-Learn