Ronok Tanvir

I am an undergraduate student studying Computer Science and Data Science at UC Berkeley. I’ve worked as a machine learning engineer intern at Halluminate and a student researcher at CAHL labs.

My experience includes designing long-horizon multi-agent environments for training with GRPO, building RL and SFT training pipelines, fine-tuning CNNs for visual detection tasks, evaluating methods for detecting LLM hallucinations, and building harnesses for containerized infrastructure.

Currently, I’m interested in AI safety, scalable oversight, and reinforcement learning, particularly understanding when automated oversight fails and how to evaluate agents beyond the signals they’re trained to optimize.