01 / 2024
RAG-based regulation policy evaluation
A retrieval system that combines language models with a dynamic knowledge base of federal regulations.
Runner-up · Stanford LLMxLaw Hackathon
Selected projects
Experiments in useful AI.
Technical work across retrieval, agent systems, evaluation, and language-model applications.
02 / 2024
Semantic SQL query assistant
A multi-agent proof of concept that generates, evaluates, and refines SQL queries through iterative prompting.
15% improvement in query correctness on a 50-task test set
03 / 2024
Fine-tuned LLMs for FAQ generation
An FAQ-generation system based on fine-tuned Llama models, designed for high-fidelity responses in constrained environments.
~0.8 BERT Score against human-written answers