Stanford, California, United States
1. Developing models with complex visual structure and scene understanding (https://arxiv.org/abs/2503.19953) 2. Developed cybersecurity benchmark and end-to-end LLM agent for evaluation (https://bountybench.github.io/)
Leading weekly labs for Stanford’s introductory CS course (CS106A/B) in Python and C++. Responsibilities include leading weekly 50-minute classes and 2-hour office hour shifts.
Implement and experiment with recent AI models, e.g. Mamba, to incorporate longer user event sequences.
Reduced recommendation model serving-time latency by 30%. Performed ablation studies and error analysis to improve model performance on CTR/CVR predictions.
Developed an interchain testing environment using Docker. Applied ACC model and wrote automated/parallel e2e and integration tests in Typescript. Shipped production-level code in Go / Typescript and performed analysis of IBC.