Ajay Mittur

Hi, I’m Ajay, a Senior Applied Research Engineer at NVIDIA, where I work on LLM post-training, agents, and agent harnesses that accelerate chip design. My work spans agent frameworks, recursive self-improvement of coding agents, and post-training large models for complex reasoning, root-cause analysis, and design verification.

I completed my MS in AI at Carnegie Mellon University in May 2025. While at CMU, I collaborated on research with Prof. Graham Neubig on LLM agents for scientific discovery — idea generation, experiment planning, code generation, and paper writing. Previously, I worked on the ARC-AGI challenge with Prof. Carolyn Rosé, researching LLM hypothesis generation and test-time learning.

Before NVIDIA, I worked with BNY on cross-LLM automatic prompt optimization, and spent the summer of 2024 on BILL’s InfoSec team building threat detection models. I’ve also worked at Amazon, Samsung R&D India, Walmart, and Honeywell across Machine Learning and Software Engineering.

Outside of work, I’m into soccer, hiking, mountain biking, and exploring :)