
I build practical RAG and ML systems that ship cleanly, feel human, and stay reliable in production.
Ship real systems, not demos.
A quick path to get the most out of this site.
A simple, repeatable process for shipping reliable ML/RAG systems and clean web products.
Define what success means, constraints, and the risk surface before writing code.
Ship a thin slice fast to validate the approach, data, UX, and feasibility.
Turn the prototype into a system with tests, observability, and repeatable deployments.
Instrument the system and improve quality where it matters: latency, accuracy, and cost.
Get it in front of users and keep it healthy. Real value comes after launch.
A few highlights that show how I work
Built a retrieval assistant for course materials with grounded, source-aware responses.
Students needed timely answers to course-specific questions outside office hours, but support bottlenecked and engagement dropped.
Built a RAG assistant grounded on syllabus + slides that returns citations and can link into multimodal sources like YouTube with timestamps.
Automated trading with sentiment analysis, technical indicators, and strict risk controls.
Automated SEO-aware content generation with GPT-4 workflows and structured output.
A few pieces worth starting with.