Kumar Krishna Agrawal

I'm a final year PhD student at EECS UC Berkeley, BAIR and UCSF, co-advised by Adam Yala and Trevor Darrell. I also work closely with Sky Computing Lab, Blake Richards at MILA/McGill.

My research is dedicated to building algorithms and systems for human-centric AI. I focus on:

  1. Advancing capabilities of AI systems: to assist clinicians worflows, focus on radiology
  2. Whiteboxing AI systems: algorithmic foundations of LLM reasoning from pre-to-post training
  3. Real-world deployment of AI systems: long-context, multimodal inference across heterogeneous compute

Before Berkeley, I was at Google Brain, where I helped build learning algorithms and systems for robot manipulation, generative models for music, and program synthesis.

I studied mathematics and computer science at IIT Kharagpur, where I was building autonomous drones, self-driving vehicles and multimodal models for video understanding. I was fortunate to spend time at MILA, advised by Yoshua Bengio.

github / gscholar