Currently working at Meta with Mike Lewis and Sharan Narang on pre-training, architectures (e.g., memory, tokenizer-free models), and data-constrained scaling.
Worked part-time for a year at Google DeepMind on modular post-training methods with Jonathan Lai, Tsendsuren, Tu Vu, and Alexandra.
Worked at Microsoft Research Redmond with Subhabrata Mukherjee and Ahmed H. Awadallah.
Worked at Amazon AWS AI Labs with Qing Sun.
Worked at Microsoft Research India with Dr. Prateek Jain before PhD.
Worked full-time for a year at LinkedIn AI Bangalore.