Hi! I'm Chetan.
I'm an engineer finishing my M.S. in EECS at UC Berkeley - graduating August 2026. I've built full-stack apps, ML and RL systems, a compiler, and embedded hardware. Currently I'm building and researching tools that help developers work more effectively with AI coding agents.
I'll be available for full-time software engineering roles from late August. Feel free to reach out at cgoenka@berkeley.edu.
Projects
RLHF for Open-Ended Instruction Following
An implementation of online policy-gradient RLHF (GRPO, DrGRPO, GSPO) for fine-tuning a 1.5B-parameter LLM on human preference pairs. The interesting parts include a replay-buffer method that builds preference pairs straight from policy rollouts, and a pessimistic reward-model ensemble that keeps the policy from gaming the reward.
Python · PyTorch · Modal · W&B
Python Compiler
A compiler I built from scratch in Java for a Python dialect. Covers the whole journey from raw source code through lexing, parsing, and type checking down to RISC-V assembly, including a simple web IDE that lets you watch each stage happen.
Java · Python · RISC-V · JFlex · CUP
Research
Think-Aloud Speech as Context for AI Code Generation
A web-based IDE that captures the way developers naturally talk through a problem and feeds it to an AI coding assistant as live context, capturing design reasoning and preferences you'd never stop to type. Three LLM agents classify the speech, keep a running model of your intentions, and decide when it's worth chiming in proactively, on top of a real-time pipeline streaming audio over WebSockets with pause-based transcript segmentation. Validated in an 11-participant user study.
Memory Bottle
An interactive bottle that captures the sensory experience of a moment and pours it back out when you tip it over at the accompanying ritual station, filling the room with the audio and light of when you saved it. Built with Arduino ESP32s and a handful of sensors that talk to a small server, keeping the audio and lights in sync.