2022 to 2024
Carnegie Mellon University
M.S. Robotic Systems Development
GPA 4.08/4.0. Visual Learning, Robot Learning, Probabilistic Graphical Models, Planning for Robotics, F1Tenth Autonomous Racing, Computer Vision.
What I am after
I want robots that get better without a human correcting every mistake. Imitation learning only teaches a policy the states a demonstrator happened to visit, so each further nine of reliability costs another pile of teleoperation. The way out is a flywheel: systems that notice when they have drifted off distribution and turn their own failures into the next batch of training data. A reactive policy is fast, cheap and confidently wrong off distribution; search is slow, expensive and able to recover. Acting Fast and Slow is putting both in one loop and spending compute only where the fast one is about to fail, which is the through-line from SAILOR to what I am building now.
How I got here
Before Nuro I worked on prediction at Plus AI and motion planning at Aurora, and before that the CMU Robotics Institute and IIT Kharagpur. Along the way I built a fair number of machines: a terrace farming robot that climbs 40 cm steps, an autonomous lunar excavator, a ground vehicle fabricated from scratch that placed twice at IGVC, and a 1/10 scale racecar that learned to drive inside a model of the world. These days it is mostly world action models, and an SO101 arm at home to keep my hands on the data and evaluation loop rather than reasoning about it from a paper.
[ 01 ]
Talks & media
Dec 2025
SAILOR, Spotlight presentation
NeurIPS 2025, San Diego
Spotlight talk and poster, top ~3% of submissions.
Sep 2025
RoboPapers, Episode 33, "A Smooth Sea Never Made a Skilled SAILOR"
Podcast · with Gokul Swamy and Arnav Jain
A full episode on world models, reward learning and test-time planning, walking through SAILOR end to end.
2025
Dec 2023
Autonomous construction on lunar-like terrain
NASA Lunar Surface Innovation Consortium (LSIC)
Presented the LunAR-X autonomous excavation work.
2022
Exchange diaries feature
IIT Kharagpur
Institute feature on the path from IIT Kharagpur to CMU.
[ 02 ]
Education
2017 to 2022
IIT Kharagpur
B.Tech Mechanical Engineering + M.Tech Systems Design, Minor in Computer Science
GPA 9.22/10. Institute Order of Merit for outstanding contributions to technology.
[ 03 ]
Awards
- 2026Runners-up, Physical AI Hack, 45+ teams, Founders Inc
- 2025NeurIPS Spotlight (top ~3% of submissions) for SAILOR
- 2025Best Self-Improving Agent, Weavehacks 2, 67 teams, W&B × Google Cloud
- 2022Invited to Research Week with Google 2022, Computer Vision
- 2022Institute Order of Merit, IIT Kharagpur, ~8 of 1500+ graduates
- 2021National runners-up, Flipkart GRID 2.0, 6500+ teams
- 2019Gold, Inter-IIT Tech Meet, best autonomous farming demonstration among 23 IITs
- 2018 & 2019Runners-up, Intelligent Ground Vehicle Challenge, Michigan, 40+ international teams
[ 04 ]
Service
- Reviewer, Conference on Robot Learning (CoRL) 2026
- Reviewer, IEEE Transactions on Robot Learning (T-RL)
- Reviewer, ICRA 2026 Workshop on Beyond Teleop
- Reviewer, RLC 2026 Workshop on RL Beyond Rewards
- Judge, Weavehacks 4, CoreWeave
- Mentor & instructor, IEEE Winter Workshop, IIT Kharagpur
[ 05 ]
Away from the bench
Two years of the National Service Scheme, a stint leading teams at AIESEC, chess since school, and enough competitive CS:GO to turn up and play it. There are photographs of most of that, along with the early robots and the people who put up with them.