Safer, efficient, and economical language models.

I research alignment, evaluation, and model behavior so LLMs can be trusted in production, run with less waste, and cost less to use well.