I build AI systems end-to-end — from data pipelines and agents to the evaluation harnesses and dashboards that prove they work. Product manager by training, engineer by habit: I've shipped ML in aerospace security, port logistics, and enterprise IT.
- MS Information Systems @ CSU Long Beach (May 2027)
- Previously: AI Product Manager @ EDRAY (port logistics) · Associate Product Manager @ Honeywell Aerospace · SWE Intern @ Dell
- Currently building multi-agent RAG systems, autonomous ops agents, and AI-native GTM tooling
| Project | What it does | Stack |
|---|---|---|
| ClauseGuard | Multi-agent RAG compliance analyst for SEC 10-K/10-Q filings — retrieval, cross-filing comparison, and a dedicated citation-verification agent so every claim traces to a source page; benchmarked with a RAGAS eval harness | LangGraph, Qdrant, ColQwen2, RAGAS |
| AI GTM Engine | End-to-end AI-native GTM pipeline: account enrichment → LLM ICP scoring → segmentation → SQL funnel analytics. Live dashboard, auto-rebuilt by CI on every push | Python, DuckDB, GitHub Actions |
| FabGuardian | Autonomous predictive-maintenance agent for semiconductor fabs — streams sensor telemetry, flags anomalies via Mahalanobis scoring, and auto-creates P1/P2 work orders with no human in the loop | watsonx Orchestrate, FastAPI, Docker |
| Ask Your Data | Plain-English → SQL for PostgreSQL, MySQL, Snowflake & BigQuery, with a query validator that blocks unsafe LLM-generated SQL | FastAPI, Streamlit, SQLAlchemy, LLMs |
| DC Arrival Prediction | Import-arrival ETA prediction over 439K+ shipment records — 1st place among 10 teams in a company-wide logistics forecasting challenge (1.89-day MAE) | XGBoost, Random Forest, scikit-learn |
| Protein-Design Studio | Single-GPU generative biology for rare-disease research: ESM-3 sequence generation + OpenFold3 structure prediction on an AMD MI300X | PyTorch/ROCm, ESM-3, OpenFold3, Streamlit |
- 75% cut in analyst processing time at Honeywell — built a Python + Neo4j graph pipeline automating cyber vulnerability assessment of embedded avionics, with 100% of findings mapped to NIST CVE/CWE/CAPEC
- 1st place of 10 teams at EDRAY — scoped and shipped an import-arrival prediction feature to production against a 1.89-day MAE launch bar
- 15% drop in incident tickets in month one at Dell — launched a self-service IAM chatbot serving a 200+ member team 24/7
RAG & vector DBs · multi-agent orchestration · LLM evaluation (RAGAS) · ETL pipelines · threat modeling & NIST mapping · Agile/Scrum