I am an Assistant Professor in the Robotics Institute and School of Computer Science at Carnegie Mellon University.

I lead the Interactive and Trustworthy Robotics Lab (Intent Lab). My research interests lie in the intersection of robotics, machine learning, and AI safety.

I obtained my Ph.D. in Electrical Engineering and Computer Sciences (EECS) from University of California Berkeley in 2022. I received a B.S. in Computer Science from the University of Maryland College Park in 2016. Before joining CMU in Sept. 2023, I was a postdoctoral scholar at UC Berkeley and spent spring 2021 working at NVIDIA in the Autonomous Vehicle Research Group.

Prospective students: please see here.

email   |   cv   |   google scholar   |   github  |   bio

Unsure how to pronounce my last name? Bajcsy sounds like BYE-chee.

• 

  [Sept 2026]

  We have 5 papers accepted to CoRL! Find the links here: StressDream, Steering via Vision & Touch, Learning What to Say to Your VLA, Do World Models Understand Partially Observable Safety Constraints?, and Scaling Test-Time Compute in Robotics.
• 

  [Aug 2026]

  I gave a talk at Georgia Tech's Robotics Colloquium!

• [July 2026]

  New arXiv paper on ELASTIC: adapting test-time compute for generative control policies.

• [June 2026]

  I gave a Keynote talk at the RPL Summer School on Robot Learning in Sweden 🇸🇪.

• [June 2026]

  New arXiv paper on ViTaL: visuo-tactile inference-time policy steering.

• [June 2026]

  New arXiv paper on WEAVER: a better, faster, and longer-horizon world model for manipulation. The pre-trained model and codebase is open-sourced too here.

• [June 2026]

  New arXiv paper on learning a language feedback policy that steers a frozen VLA; the policy learns what to say to the VLA and (crucially) when not to say anything.

• [June 2026]

  New arXiv paper on StressDream: a noise optimization algorithm for steering pre-trained video world models. Check out how StressDream can help with robust policy evaluation and improvement here.

• [May 2026]

  New arXiv paper on Coordinated Diffusion: an algorithm for generating multi-agent robot behavior without multi-agent demonstrations.

• [May 2026]

  I gave a talk at the University of Washington's Robotics Colloquium!

• [April 2026]

  Our position paper on reward modeling was accepted as a spotlight to ICML (top 5%)!

• [April 2026]

  Our paper on uncertainty-aware policy steering was accepted to RSS!

• [March 2026]

  New arXiv paper on uncertainty-aware policy steering.

• [Jan 2026]

  Three papers accepted to ICRA! Check out the publications on the lab website.

• [Jan 2026]

  One paper on latent CBFs accepted to L4DC as an oral talk!

[Older News]

[Dec 2025]

One paper on control-theoretic LLM guardrails accepted to the International Association for Safe and Ethical AI!

[Nov 2025]

New arXiv paper on latent-space control barrier functions for visuomotor policy safety.

[Nov 2025]

Honored to receive the Amazon Research Award!

[Oct 2025]

I gave a talk at IVADO Workshop on Assessing & Improving Agent Capabilities & Safety. Titled From Refusal to Recovery: Control-Theoretic Approach to Agent Guardrails and you can check it out on YouTube!

[Oct 2025]

Four new arXiv papers! We studied how latent-space safe control can go wrong under partially observable constraints but multimodal data is a promising remedy, how to adapt latent safety filters at runtime with images of constraints, how we can automatically compute agentic AI guardrails by performing control-theoretic optimization directly on LLM agent models, and how we can steer VLAs at runtime by verifying that their intermediate reasoning steps actually align with the outcomes of their actions.

[Sept 2025]

I'm honored to receive the DARPA Young Faculty Award (YFA)!

[Aug 2025]

Two papers accepted to CoRL! One on controlling away from known and out-of-distribution failures and one on out-of-distribution adaptation at runtime via "analogy".

[May 2025]

Check out the Robotics Institute's article about my CAREER award and our initial results.

[May 2025]

New paper on arXiv on uncertainty-aware latent safety filters which proactively steer robots away from both known and out-of-distribution failures.

[May 2025]

We ran a successful 2nd edition of the Northeast Systems and Control Workshop!

[April 2025]

Honored to receive the NSF CAREER Award!

[April 2025]

Two papers accepted at RSS! Check them out here and here.

[March 2025]

Check out my talk titled "Towards Open World Robot Safety" at the RI Seminar!

[Feb 2025]

Two new arXiv papers: latent-space reachability for generalizing safety beyond collision-avoidance and VLM-in-the-loop policy steering!

[Jan 2025]

Two accepted papers (updating safety via language and safely influencing) at ICRA 2025!

[Jan 2025]

One accepted paper on conformalized IL at ICLR 2025!

[Jan 2025]

We are running two workshops at ICRA! One on Safely Leveraging VLMs in Robotics and the other on Public Trust in Autonomy

[Oct 2024]

New arXiv paper on conformal prediction + interactive imitation learning.

[Sept 2024]

Three new arXiv papers! Check out our work on robots that suggest safe alternatives, updating robot safety representations online from natural language, and robots that learn to safely influence.

[Sept 2024]

New paper accepted to CoRL on how not all prediction errors are made equal from the perspective of closed-loop robot performance.

[Aug 2024]

New paper at ECCV on adapting human trajectory predictors to scene-specific behaviors.

[June 2024]

New arXiv paper on human--AI safety through the lens of generative AI and control systems.

[Apr 2024]

New paper at RSS on using conformal prediction to calibrated shared controllers.

[Apr 2024]

I'm thrilled to receive the Google Research Scholar Award!

[Mar 2024]

New arXiv paper on a general calibrated regret metric for detecting and mitigating human-robot interaction failures; check out the applicability to anomaly detection in interactive generative planners too!

[Feb 2024]

New arXiv paper on intent demonstration in general-sum dynamic games.

[Jan 2024]

Our work on visual representation alignment for robot learning was accepted to ICLR 2024! Check out the new results in the arXiv version.

[Jan 2024]

Our work on contingency games was accepted to RA-L 2024!

[Jan 2024]

Our work on stabilized and robust online learning from humans was accepted to RA-L 2024!

[Dec 2023]

New pre-print on Adaptive Human Trajectory Prediction via Latent Corridors.

[Oct 2023]

Submitted a paper to ICLR! We propose Representation-Aligned Preference-based Learning (RAPL), a tractable video-only method for solving the visual representation alignment problem and learning visual robot rewards via optimal transport.

[Sep 2023]

Submitted a paper to ICRA! We present Conformal Decision Theory, a new theoretical and algorithmic framework for online calibration of decision risk.

[Aug 2023]

ArXiv paper on learning vision-based pursuit-evasion robot policies. Check out our project website for videos of human-robot and robot-robot interaction ``in the wild''.

[Aug 2023]

1 paper accepted to CoRL! This work synthesizes safe control policies that explicitly account a robot's ability to learn and adapt at runtime.

[Aug 2023]

Submitted a paper to RA-L on stabilized and robust online learning from humans.

[Aug 2023]

Extended results on contingency games now on arXiv.

[Aug 2023]

I gave a talk at Bosch research.

[Apr 2023]

New paper on contingency games: a model for strategic interactions which allows a robot to consider the full distribution of other agents’ intents while anticipating intent certainty in the near future.

[Feb 2023]

I gave a talk at the Illinois Robotics Seminar at UIUC.

[Feb 2023]

I gave a talk at the Wisconsin Robotics Seminar Series at the University of Wisconsin-Madison.

[Jan 2023]

Our paper on long-term robot influence on humans got accepted to ICRA 2023!

[Jan 2023]

Our paper on modeling & influencing the dynamics of human learning was accepted to HRI 2023!

[Dec 2022]

I am on the organizing committee of RSS 2023.

[Dec 2022]

I gave a guest lecture at Princeton's Safety-Critical Robotic Systems class.

[Nov 2022]

I am serving as an Area Chair for L4DC.

[Oct 2022]

I gave an invited talk at Nuro on practical safety for human-robot interaction.

[Sept 2022]

Paper submitted to ICRA on robots that influence humans over long-term interaction.

[Sept 2022]

I'm serving as Associate Editor for ICRA 2023.

[Aug 2022]

I completed my PhD and started my postdoc! My thesis entitled "Bridging Safety and Learning in Human-Robot Interaction" can be found here.

[July 2022]

I gave an invited talk at the ICML Workshop on Safe Learning for Autonomous Driving (SL4AD).

[June 2022]

I gave an invited talk at the RSS Workshop on Close Proximity Human-Robot Collaboration.

[May 2022]

I accepted an offer to join the Robotics Institute in the School of Computer Science at Carnegie Mellon University (CMU) starting Fall 2023!

[Mar 2022]

Our paper Safety Assurances for Human-Robot Interaction via Confidence-aware Game-theoretic Human Models was accepted to ICRA 2022!

[Nov 2021]

I gave a talk at the University of Washington's Robotics Colloqium.

[Oct 2021]

Paper published in IJRR! In this work we formalize how physical human interaction with a robot can be treated as a form of communication, enabling robots to learn objectives from human corrections.

[Oct 2021]

New paper on arXiv! In this work we propose that robots use confidence-aware game-theoretic models of human behavior when assessing the safety of a human-robot interaction.

[Aug 2021]

I am grateful to have been selected for the Rising Stars Academic Career Workshop in EECS.

[Aug 2021]

New paper on arXiv! In this work we advocate for the use of Hamilton Jacobi (HJ) reachability as a unifying mathematical framework for comparing existing safety concepts used throughout industry and academia, and propose ways to expand its modeling premises in a data-driven fashion.

[April 2021]

I'm excited to join NVIDIA for an internship in the Autonomous Driving Research group led by Marco Pavone.

[Mar 2021]

Our paper on analyzing human models that adapt online was accepted to ICRA 2021!

[Feb 2021]

I gave a talk at University of Pennsylvania on analyzing human models that adapt online.

[Dec 2020]

I gave a talk on safe robots which learn from (and about) humans at the University of Chicago Laboratory School for the middle and high-school robotics club.

[Dec 2020]

I gave a talk at ETH Zurich on introspective human motion prediction for safe robot autonomy.