Manik Bhandari

I’m Senior Research Scientist at Meta Superintelligence Labs where I work on Reinforcement Learning, LLM post-training, Safety, and Alignment. My recent work focuses on training better reward models (presented at ICML 2026) and developing LLM post-training methods that balance Safety and Instruction Following.

Previously, I was an Applied Scientist at Amazon, working on Search and Recommender Systems to improve Alexa’s Speach Recognition and Language Understanding. I did my Master’s at Carnegie Mellon University advised by Prof. Pengfei Liu and Prof. Graham Neubig where I worked on text summarization. Prior to that, I was a research assistant at Indian Institute of Science advised by Prof. Partha Talukdar. My research broadly focuses on LLMs, Safety Alignment, Search, Recommender Systems, Knowledge Graphs and Graph Convolutional Networks (GCNs).