Gabriel Cha
i like to think, explain, build.
agentic harnesses that reason, search the web, and synthesize information fascinate me. so does fine-tuning llms on a100s to push quality on narrow tasks.
Paper
-
Accepted to ACM SIGCSE 2026An experience report on prototyping a JupyterLab extension that generates programming practice questions inside the instructor's notebook — evaluated through evidence-based prompt engineering and a usability study with six data science instructors.
Selected Works
-
An open-source CLI with a local LLM harness — runs on Ollama or Claude, retrieves relevant code via a CodeBERT + BM25 RAG, and proposes diffs for failing commands.
-
Awarded UCSD HDSI Research Scholarship 202412,000 clips across RAVDESS, TESS, CREMA-D, and SAVEE — reaching 86% test accuracy with a CNN. Funded by UCSD's $6,500 HDSI Undergraduate Research Scholarship.
-
Estimating sequence likelihoods from word counts and smoothing — no neural networks.
