Prakhar Agarwal

I am a Researcher specializing in LLM post-training and alignment, currently working at Meta Superintelligence Labs.

At Meta, I focus on enhancing the long-context reasoning capabilities of LLMs. My work involves extending instruction following, factuality, and efficient tool use for 50+ dialects—directly impacting over 600 million users globally. I am responsible for improving model response quality by developing rigorous evaluations and hill-climbing them using state-of-the-art post-training techniques, including mid-training, SFT, LLM-judge/BT-RM training, rubric-based RL, and distillation.

Before joining Meta, I worked as a Member of Technical Staff at OpenAI, and spent five years as a Machine Learning Engineer at Apple. I hold a graduate degree specializing in Machine Learning from the University of Washington, and a Bachelor's from BITS Pilani.

News & Media

Business Insider ↗
Discussing the reality of working at frontier AI labs and the fast-paced, high-autonomy culture.

The Economic Times ↗
Thoughts on hiring practices at top AI labs and the importance of practical problem-solving.

NDTV ↗
Sharing advice on how top AI labs prioritize hands-on experience over traditional academic credentials.