MCP tools for retrieval across teams
Model Context Protocol client-server tools enabling RAG workflows across teams while preserving privacy boundaries.
Engineering and applied ML work from industry roles and coursework.
Model Context Protocol client-server tools enabling RAG workflows across teams while preserving privacy boundaries.
Evaluation pipelines measuring transcription accuracy across 10,000+ hours of Apple Music Radio interviews.
Neural forecasting pipelines for statewide demand at 2% error, Ameren's first internal forecasting tool, with a projected $12M annual savings in MISO trading.
Simulated production at proposed solar and wind sites from historical weather to validate geographic diversity for regulators.
With Grace Pfeiffer and Aden Weiss, and in partnership with the nonprofit Caras Con Causa, co-designed an ebb-and-flow test stand for growing mangroves in bulk under controlled water level, salinity, and circulation, for student experiments at a school in Guaynabo, Puerto Rico, and traveled to Puerto Rico to work with Caras on site. Also designed the sensor-based pump control system (Raspberry Pi feedback loop), which REU students later built; a proposed reinforcement-learning controller was set aside for a simpler setup the partner's staff could maintain.