Now
AI agents for education data
Building an agent that understands the education data standards our clients actually run on — Ed-Fi, TEA TSDS, College Board, Accuplacer — and answers our users' questions directly from their own data. I am training it now, and teaching it to educate itself on how the Texas and California education systems work.
Now
Agentic migration off the legacy codebase
Running an NVIDIA Spark box in the office as on-prem inference for a two-tier agent system: Claude orchestrates the migration plan, and a local Qwen model does the heavy code refactoring. The goal is to finally retire the legacy codebase and run entirely on our modern framework.
2016 —
Mission-critical state data pipelines
Architected the real-time synchronization engine for the Texas Education Agency API, then extended the same fault-tolerant ingestion architecture to California's Aeries SIS. Over 1.1 billion rows moved at full state reporting compliance, with no data latency.
Ongoing
React and TypeScript across a Laravel platform
Our product runs on PHP and Laravel, and still does. My work has been building its React and TypeScript surface and the Node scripting around it — and I am one of two developers on the team who can write in those languages without leaning on AI to do it.