Nikhil Dhawan

Research Interests

Reinforcement LearningAgentic Software EngineeringLarge Language ModelsMachine LearningCausal InferenceEconometricsEvidence-Based Medicine

Hello, I'm

Nikhil Dhawan

Senior Researcher at Google DeepMind

I'm currently a researcher at Google DeepMind working on reinforcement learning and post-training to improve Gemini's coding & software engineering capabilities.

Prior to DeepMind, I was a software engineer at Google working on ads optimization and recommendations. Before that, I was a software engineer and data scientist at Microsoft, working on research economics and machine learning problems related to Microsoft Azure. During my time there, I had the great fortune of working with and reporting to the Senior Director of Economics in Azure, Patrick Hummel. I've also previously interned at The Blackstone Group and as a researcher at Cornell Tech.

I attended Cornell University from 2015–2019 where I studied Computer Science, Electrical Engineering, and Business. While at Cornell, I was an undergraduate researcher in the Human Robot Collaboration and Companionship Lab under the supervision of Guy Hoffman. My research focused on studying side-by-side collaborative design between humans and intelligent agents, and I co-authored several publications including one which won Best Paper at the 2018 Design Computing and Cognition conference.

I live in New York City and grew up in the Hudson Valley. I also enjoy cooking—I've even taken classes at the Culinary Institute of America—lifting weights, and eating dinner at Don Angie.

NYC Skyline Cooking San Diego Sunset Don Angie

Publications

2025

Sulfur-or Heavy Atom-Containing Nanoparticles, Methods of Making Same, and Uses Thereof

F.F.E. Kohle, J.A. Hinckley, N. Dhawan, U.B. Wiesner

US Patent 12,397,067

View Publication
2021

Hammers for Robots: Designing Tools for Reinforcement Learning Agents

M.V. Law, Z. Li, A. Rajesh, N. Dhawan, A. Kwatra, G. Hoffman

ACM Designing Interactive Systems Conference (DIS)

This paper discusses the design of tools tailored for reinforcement learning agents, enhancing their interaction capabilities and enabling more effective human-robot collaboration.

View Publication
2020

Design Intention Inference for Virtual Co-Design Agents

M.V. Law, A. Kwatra, N. Dhawan, M. Einhorn, A. Rajesh, G. Hoffman

ACM International Conference on Intelligent Virtual Agents (IVA)

Focuses on inferring design intentions to improve collaboration between humans and virtual co-design agents, enabling more intuitive and effective design processes.

View Publication
2019

Amorphous Quantum Nanomaterials

F.F.E. Kohle, J.A. Hinckley, S. Li, N. Dhawan, W.P. Katt, J.A. Erstling, et al.

Advanced Materials

Introduces amorphous quantum nanomaterials with potential applications in various technological fields, exploring novel material properties and synthesis methods.

View Publication
2018

Side-by-Side Human–Computer Design Using a Tangible User Interface

M.V. Law, N. Dhawan, H. Bang, S.Y. Yoon, D. Selva, G. Hoffman

International Conference on Design Computing and Cognition (DCC) — Best Paper Award

Best Paper Award. Presents a tangible user interface facilitating collaborative design processes between humans and computers.

View Publication
2019

Preventing Catastrophic Forgetting in an Online Learning Setting

A. Borthakur, M. Einhorn, N. Dhawan

Research Paper

Explores methods to prevent catastrophic forgetting in neural networks during online learning scenarios.

View Publication

Projects

Bluetooth-Enabled Buggy

A two-wheeled differential drive robot that is controlled via a moblile bluetooth terminal. We wrote a C program which interfaced with an NXP development board to receive UART communications and output PWM signals to each motor. We also deployed a python script onto a raspberry pi which read incoming serial messages from bluetooth and relayed it to a the NXP board.

CPythonBluetooth

PokerBot

A text-based Poker Game and Poker AI in OCaml. During each round, the bot ran hundreds of simulations taking into account the current river, user's betting strategy, pot odds, etc. and made a decision to call, raise, check, or fold.

OCaml

Timeline

Senior Researcher

Google DeepMind

2025 — Present

Working on Gemini, focusing on reinforcement learning and post-training to improve coding abilities and software engineering capabilities.

Software Engineer

Google

2021 — 2024

Worked on building LLMs for coding in Google Labs. Previously worked on ads optimization and recommendations, developing machine learning models for improved ad targeting.

Software Engineer / Data Scientist

Microsoft Azure

2019 — 2021

Worked on research economics and machine learning problems for Azure cloud compute. Reported to Patrick Hummel, Senior Director of Economics.

Software Engineering Intern

The Blackstone Group

Summer 2018

Interned at one of the world's leading investment firms.

B.Sc. Computer Science

Cornell University

2015 — 2019

Studied Computer Science, Electrical Engineering, and Business. Undergraduate researcher in the Human Robot Collaboration and Companionship Lab under Guy Hoffman.

High School

Spackenkill High School

2011 — 2015

Salutatorian. President of Science Olympiad. Co-captain of the Math Team and Mock Trial.