AI/ML Engineer · I build multi-agent systems and LLM platforms that actually ship.
(yes, the agents sometimes disagree with each other. that's a feature.)
At CBRE, I built a Palantir AIP multi-agent appraisal platform serving 600+ appraisers across Wells Fargo, Bank of America, and Bank OZK cutting processing time from hours to 30 seconds.
At JPMorgan Chase, I co-invented U.S. Patent 12,223,309 for an ML-driven SDK upgrade detection system, and spent ~2.5 years building distributed Java systems that (mostly) didn't page anyone at 3am.
M.S. in Artificial Intelligence, Northeastern University (Khoury College, Dec 2025). Also TA'd CS 4100 and CS 5047 there — turns out explaining AI to students is a great way to find out what you don't actually understand.
Planner → Generator → Debugger pipeline with PPO-based reward model training and human/AI feedback loops. The RLHF loop uses human preference labels + GPT-4 as a judge to iteratively improve code quality. Three agents arguing about your code so you don't have to.
LangGraph · Qwen2.5 · RLHF / RLAIF · PPO · Streamlit
Semantic document querying with chunked embeddings, FAISS vector retrieval, and conversational memory. Handles multi-document context with source attribution — because "I read it somewhere" isn't good enough for an LLM either.
LangChain · FAISS · OpenAI · Python
Natural language to functional webpage pipeline — parses intent, maps to layout components, renders HTML/CSS end-to-end. Still faster than most Figma handoffs.
Python · Transformers · NLP
The portfolio projects I work on at night after the agents at work finish arguing.
- LLM Evaluation Harness — async job queue (Redis/RQ), PostgreSQL result store, RAGAS / DeepEval / BERTScore scorers, GPT-4o as judge. Because vibes are not a benchmark.
- Multi-Agent ICU Deterioration Monitor — MIMIC-IV data, AWS MSK Kafka streaming, LangGraph orchestration, ECS Fargate. High stakes, low tolerance for hallucinations.
| Languages | Python · SQL · JavaScript |
| Agentic AI | LangGraph · LangChain · OpenAI Agents SDK · Palantir AIP |
| ML / DL | PyTorch · HuggingFace · Scikit-learn · RLHF / PPO |
| Infrastructure | Kafka · AWS (MSK, S3, EC2) · Docker · FastAPI · PostgreSQL |
| Vector DBs | FAISS · Pinecone · Weaviate |
- 🏫 Teaching AI at Northeastern made me realize most AI problems are actually communication problems
- 🧪 Currently obsessed with LLM evaluation — turns out measuring intelligence is harder than building it
- 🚂 Strong preference for train-accessible day trips (Newport is underrated)
- 📬 Ask me about: multi-agent orchestration, Palantir AIP, RLHF pipelines, or why your RAG isn't working
Open to full-time AI/ML Engineer roles. Let's connect →