Jason Trinh

Home

Jason Trinh

I'm an EECS student at UC Berkeley interested in machine learning, systems, optimization, and probability.

Relevant Coursework

Deep Learning

CS 182/282A · A+

Convex Optimization

EE 227BT · A

Reinforcement Learning

CS 185/285 · A

Scalable AI

EE 194/294 16 · A+

View full coursework

Research

EchoKV NeurIPS 2026 Submission

Adaptive per-layer KV-cache precision for large language model serving.

Reward-Filtered Policy Optimization

Improving DPO-style preference optimization with reward-filtered sampling.

Deep Learning: Riding the Polar Express to Convergence

Analyzing maximum stable learning rates for optimizers.