I build generative-AI systems that reach production — real-time LLM tools, RAG, and NLP at scale — and lead the teams that ship them.
Flagship generative-AI systems I've taken from prototype to production at Capital One — starting with a public, clickable rebuild you can try yourself.
A public rebuild of the shape of the three systems below, running live in your browser. Talk into your mic — or play a sample call — and watch a two-voice transcript, self-drafting notes, RAG-retrieved procedure docs, and a sub-second frustration alert, all from a single WebSocket. Deepgram + OpenAI composed into a real-time product; no proprietary anything.
The problems I love solving.
Agentic LLM pipelines, RAG, and prompt/eval systems that augment thousands of users live, with measurable business outcomes.
Sentiment, complaint, and intent detection over streaming call transcripts — with sub-second alerting for de-escalation.
LLM-as-a-judge, offline & online evaluation, and observability for non-deterministic systems in production.
Low-latency, real-time streaming inference on Kafka, AWS Lambda, DynamoDB, and Snowflake.
Auto-loan underwriting models decisioning loans at large scale, plus reinforcement-learning credit-policy optimization.
Leading ML engineers and data scientists, setting technical objectives with VP- and EVP-level stakeholders.
Taking models to production — tap a skill to read the case study.
I'm a Machine Learning Engineer and Data Science Manager at Capital One, where I lead Generative AI for the contact centers — LLM-powered, agentic tools that augment thousands of servicing agents live, mid-conversation.
Over six years I've taken applied AI from prototype to production: real-time NLP, RAG, low-latency streaming inference, and the credit-risk ML that decisions loans at large scale.
What drives me is making a product measurably better for the people using it. Outside work: science fiction (Dune above all), strategy games, and time with my wife Amy and our dog Huck.
