Building production AI systems that solve real business problems.

Caldwell, NJ

• Built a production agentic RAG copilot converting AML/sanctions alerts into examiner-ready case files with grounded evidence citations — requiring explainability, auditability, and high citation coverage in a regulated environment.
• Designed multi-agent orchestration pipeline combining retrieval, tool-use, and structured output templates; collaborated with BSA/AML officers and ACAMS NJ Chapter to define evidence standards and reviewer acceptance metrics.
• Engaged mid-sized banks (Valley, OceanFirst, Citi, Columbia) for 90-day pilots; commercial pipeline targets $50K–$70K ARR per institution.

Caldwell, NJ

• Ran rapid POCs for property history retrieval, tracking, and insight scoring; shipped a paid subscription product iterated on real usage signals.
• Selected for LAUNCH Founder University (Jason Calacanis); raised $5K pre-seed. Achieved early revenue with 50+ paid homeowner subscriptions and 10+ agent annual subscriptions.
• Productionized platform with API-first backend and analytics pipelines; improved relevance through usage-driven iteration.

Dublin, TX

• Deployed RL-based NPC behavior and difficulty scaling with production constraints; achieved p95 inference latency ≤ 12ms and maintained 60 FPS on target hardware.
• Optimized GPU inference and reduced model size from 120MB → 45MB (-62%), improving load time by ~30% and reducing runtime memory usage by ~25%.
• Built an evaluation harness for NPC behavior quality (win-rate balance, difficulty curve, policy regressions) and reduced unfair-difficulty incidents by ~35%.

Caldwell, NJ

• Implemented 10+ CNN image detection models in Python/TensorFlow/Keras; improved non-backprop performance to 98% accuracy via transfer learning and filter pruning.
• Built activation-based filter update/pruning workflow; improved CNN accuracy by 5.67% and reduced training time by ~30.5 seconds.

Caldwell, NJ

• Shipped backend features for a production web platform, building API-first services and maintaining clear technical documentation to support ongoing iterations and releases.
• Implemented 20+ automated test cases across unit, integration, and end-to-end suites to improve reliability and prevent regressions during deployment.
• Delivered a responsive database management web app used by 8,000+ university users; integrated authentication, data workflows, and REST APIs backed by Firebase to support secure personal/public data operations.