Neil Pagarkar Rathi

Hi, I'm Neil. I study mathematics and ancient history at Stanford.

Right now, I mostly think about how to make sure that language models are good for society. I'm currently a resident on Alignment at Anthropic. I also work with Dan Jurafsky in the Stanford NLP Group.

I'm also interested in postcolonial theory and the California coast. I consume a lot of media.

You can reach me at lastname [at] stanford [dot] edu.

Some things I've done recently

• Shaping model capabilities by filtering pretraining data
• Alternative evaluation paradigms for language models
• Simple train-time interventions to reduce bad generalization
• Logistics for the largest high school math tournament in the country
• Archival work on colonialism and caste in the India Office Records

scholar / twitter / last.fm