Raghavendra Kotikalapudi

Raghavendra (Ragha) Kotikalapudi

I am an AI researcher at Microsoft AI on the MSI team, working on multi-step reasoning and reinforcement learning.

Previously at Google DeepMind, I was a key contributor to Gemini's post-training efforts across most model releases—from Gemini 1.5 Pro through 2.5 Pro and thinking models. As one of the leads of the ICPC team, I contributed to Gemini winning ICPC gold—achieving superhuman competitive programming performance (2nd place overall, solving 10/12 problems including one no human team solved). I was also the research lead for instruction tuning and safety on Bard and Gemini 1.5, and was part of the small strike team that launched the initial version of Bard in 100 days.

Media

  • Jun 2026 Gemini STOC Reviewer — Contributed to the Gemini Deep Think team that provided automated feedback for STOC 2026. Over 80% of submitted papers opted in, with 97% finding the feedback helpful.
  • Sep 2025 Gemini ICPC Gold — Key contributor to the Gemini 2.5 Deep Think team that achieved gold-medal level performance at the ICPC World Finals in Baku. Placed 2nd overall, solving 10/12 problems including Problem C which no human team could solve.
  • Oct 2017 Facebook Hackathon Winner — Won the Facebook Civic Hackathon with "Find 'n Park", a computer vision model to detect available parking spots in real-time using Seattle's open data.
  • Jun 2012 NYTimes Op-Ed — Co-authored "How Depressives Surf the Web" in the New York Times Sunday Review, discussing research linking internet usage patterns to depression in college students.

Selected Publications

Full list on Google Scholar

Patents