Roozbeh Mottaghi

Roozbeh Mottaghi
Member of Technical Staff
Skild AI

Affiliate Associate Professor
University of Washington · School of Computer Science and Engineering




Leading research on Embodied AI, focusing on building AI systems that can perceive, reason, and interact with the physical world.

Featured Work

Embodied AI Simulators

AI2-THOR/ProcTHOR

Large-scale scene generation for Embodied AI.

NeurIPS Outstanding Paper Award
Habitat 3.0

Simulator for humans and robots.

World Models

ODIN

Learning to imagine the 3D world.

ObjectForesight

Predicting future 3D states of objects.

Planning & Reasoning

PARTNR

A benchmark and suite of models for embodied planning and reasoning.

Mark Zuckerberg Announcement
Embodied Learnable Memory

Planning with a dynamic memory.

Robot Navigation

Target-driven Navigation

RL for indoor navigation.

Go To AnyThing (GOAT)

Robot navigation generalizable to different AirBnBs.

Object Manipulation

HomeRobot

Open-vocab mobile manipulation.

Track2Act

Learning manipulation from internet videos.

Physical Reasoning

Newtonian Image Understanding

Learning dynamics of objects in images.

InfLevel

Benchmarking progress to infant-level physical reasoning.

Self-Adaptive Models

SAVN

Self-adaptive robot navigation.

Interactron

Self-adaptive object detection.

Representation Learning

DECADE

Learning from dog behavior.

MuscleTorch

Learning from muscle interactions.

Reasoning-based VQA

OK-VQA

VQA with external knowledge.

A-OKVQA

VQA with world knowledge.

View all publications →

Recent Updates

Nov 2025 Giving invited talks at the Holistic Video Understanding and Vision Language Models: Challenges of Real-World Deployment workshops at NeurIPS 2025. I will talk about generating and understanding the 3D world and human-agent collaborative planning.
Nov 2025 Giving an invited talk at the CSE Colloquium at UCSC. I will talk about mitigating data scarcity in robotics via simulation.
Nov 2025 Serving as the lead Area Chair for CVPR 2026 and Area Chair for ICLR 2026.
Oct 2025 Giving invited talks at the Human-aware Embodied AI and AI Meets Autonomy: Vision, Language, and Autonomous Systems workshops at IROS 2025, and Human-Robot-Scene Interaction and Collaboration workshop at ICCV 2025.
Jun 2025 Giving an invited talk at the 3D-LLM/VLA: Bridging Language, Vision and Action in 3D Environments and Embodied Humans: Symbiotic Intelligence between Virtual Humans and Humanoid Robots workshops at CVPR 2025.
Apr 2025 Giving an invited talk in VCR/AI seminars at Simon Fraser University.

Older updates →

Background

I am a Member of Technical Staff at Skild AI and an Affiliate Associate Professor in Paul G. Allen School of Computer Science & Engineering at the University of Washington.

Previously, I was a Senior AI Research Scientist Manager at Meta FAIR and a Research Manager on the PRIOR team at the Allen Institute for AI. I also held a postdoctoral position at Stanford University. I received my Ph.D. from UCLA, advised by Alan Yuille.

Students & Interns

I have had the privilege of working with the following PhD students, residents, and interns.

Kiana Ehsani PhD student (2016–2021), Co-advisor → Research Scientist at AI2
Kuo-hao Zeng PhD student (2018–2023), Co-advisor → Research Scientist at AI2
Mitchell Wortsman AI2 PYI (2018–2019) → PhD student at U. Washington
Klemen Kotar AI2 PYI (2020–2022) → PhD student at Stanford
Apoorv Khandelwal AI2 PYI (2020–2022) → PhD student at Brown
Kunal Singh AI2 PYI (2021–2023) → PhD student at EPFL
Homanga Bharadhwaj Meta AIM (2023–2024) → Research Scientist at Meta FAIR

Interns

Yuke Zhu Stanford Daniel Gordon U Washington Kenneth Marino CMU Arun Mallya UIUC Fan Yang CMU Prithvijit Chattopadhyay Georgia Tech Vishvak Murahari Princeton Qian Huang Cornell Samir Gadre Columbia Kshitij Dwivedi Goethe Amanda Rose Yuile UIUC Peng Gao CUHK Jialin Wu UT Austin Karl Schmeckpeper U Penn So Yeon Min CMU Ram Ramrakhya Georgia Tech Cem Gokmen Stanford