San Francisco, California, United States
Life-longer learner
Doing research
Azure OpenAI evals: GPT-5, Deep Research, Sora, Sora 2, o3, o4-mini, codex-mini, and more!
TA for 10623 Generative AI (Grad Level) Translated recent research ideas into stable, production-quality assignments, iterating under real student feedback and ambiguity; awarded TA Award (Fall 2024) for sustained impact
TA for 10601 Machine Learning Led the redesign of core ML coursework, modernizing assignments to neural language modeling and Transformers in PyTorch and designing end-to-end systems that require implementing multi-head self-attention, pretraining, and downstream fine-tuning to emphasize representation learning
• Fine-tuned and evaluated multiple foundation models (Llama 3, Phi-3, Mistral) on a custom-curated speech recognition dataset, addressing systematic failure modes including mispronunciations, out-of-distribution names, and accented speech through data curation and training strategy iteration • Applied state-of-the-art instruction tuning techniques on GPT-4o, designing prompt and supervision strategies to elicit consistent, controllable behavior on speech-centric tasks under limited labeling and deployment constraints