Mar 2024 – Oct 2024 · 7 mos

• RLHF
• Lead one of Gemini Pro launches significantly improving reasoning/code and the first launch to successfully incorporate user feedback in the RL recipe.
• RL captain for several releases of Gemini free tier models.
• Improved stability of RL recipe for Gemini basic models.
• Reward model training
• Release owner of Reward model for Bard advanced launch.
• RM owner for Gemma 2 27B launch.
• Worked with external vendors on collecting Human labeled preference data for RM training for Gemini 1.0 and 1.5 family of models.

Mar 2019 – Mar 2024 · 5 yrs

• Natural Language Generation service to produce natural and fluent output across languages.
• Infra for integrating with text editing/repair models significantly reducing the time/effort spent authoring templates.
• WebAnswers
• Offer informative punts to the user based on extracted intent to better clarify why the query could not be answered. This improved user satisfaction and led to reduction in repeated queries.
• Improve user interaction by offering follow-up for information seeking queries providing them with a more detailed answer if the user accepts the suggestion.

Mar 2017 – Jan 2019 · 1 yr 10 mos · Bengaluru Area, India · On-site

• Building large scale, real-time streaming platform for dynamic pricing

Jun 2015 – Mar 2017 · 1 yr 9 mos · Bengaluru Area, India

Jan 2015 – Jun 2015 · 5 mos · Bengaluru Area, India

Dec 2013 – Jan 2014 · 1 mo · Hyderabad Area, India