Parsa Mahmoudieh

I am currently a researcher in Gemini at Google Deepmind based in Mountain View. My most recent contributions has been in the RM & RL stages of Gemini 2.5 and Gemini 2.0. I also have co-led the RM & RL stages for LearnLM which were presented at Google I/O 2024 and Google I/O 2025.

Previously I received my CS PhD at UC Berkeley in BAIR advised by Trevor Darrell and have been mentored by Evan Shelhamer and Deepak Pathak. During my PhD I mostly worked on Self-Supervised Reinforcement learning and Behavior Cloning.

Before grad school, I did a double major in EECS and MechE at UC Berkeley and had done undergraduate research in Ron Fearing's Robotics lab. I've also had the pleasure to do internships at GM and Ford.

Outside of sand volleyball and basketball, one big side hobby of mine is studying physiology, molecular biology, and computational biology.

Email  /  Google Scholar  /  LinkedIn  /  PhD Thesis

Research
Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities
One of the core contributors, 2025
[Blog] [Tech Report]

Evaluating Gemini in an Arena for Learning
One of the core contributors, 2025
[Blog] [Tech Report] [arXiv]

LearnLM: Improving Gemini for Learning
One of the core contributors, 2024
[Blog] [Tech Report] [arXiv]

Towards Responsible Development of Generative AI for Education: An Evaluation-Driven Approach
One of the core contributors, Google I/O 2024
[Blog] [Tech Report] [arXiv]

Zero-Shot Reward Specification via Grounded Natural Language
Parsa Mahmoudieh, Deepak Pathak, Trevor Darrell
International Conference on Machine Learning (ICML), 2022   (Spotlight talk)

Weakly-Supervised Trajectory Segmentation for Learning Reusable Skills
Parsa Mahmoudieh, Trevor Darrell, Deepak Pathak
International Conference on Learning Representations (ICLR) Workshop, 2020

Zero-Shot Visual Imitation
Deepak Pathak*, Parsa Mahmoudieh*, Guanghao Luo*, Pulkit Agrawal*, Dian Chen, Fred Shentu, Evan Shelhamer, Jitendra Malik, Alexei A. Efros, Trevor Darrell (* equal contribution)
International Conference on Learning Representations (ICLR), 2018   (Oral Presentation)

Loss is its own Reward: Self-Supervision for Reinforcement Learning
Evan Shelhamer, Parsa Mahmoudieh, Max Argus, Trevor Darrell
International Conference on Learning Representations (ICLR) Workshop , 2017

Modeling and Control of an Ornithopter for Diving
Cameron J. Rose, Parsa Mahmoudieh, Ronald S. Fearing
International Conference on Intelligent Robots and Systems (IROS) , 2016

Coordinated Launching of an Ornithopter with a Hexapedal Robot
Cameron J. Rose, Parsa Mahmoudieh, Ronald S. Fearing
International Conference on Robotics and Automation (ICRA) , 2015


Template from Jon Barron