Shahram Najam Syed

Robot learning · World models · VLAs · Perception · Action segmentation · Manipulation

Shahram Najam Syed

I make robots (and the models inside them) go brrrr.

Robot learning should inherit its architecture from biological sensorimotor control. Nervous systems never wait for a world that holds still. They meet uncertainty with one unbroken cascade from perception to memory, and that cascade is what lets a body act in places it has never been. I want robots to learn the same way, closing the loop between language, vision and action so that every action becomes something they understand.

  1. 01 Segment

    The brain binds the senses and carves continuous experience into discrete events. Multimodal temporal representations align video in time, segment long-horizon procedures with little or no supervision, and let touch stand in when vision drops out.

    Aligning Videos in Time · Visuo-Acoustic Hand Pose · THING (in progress)

  2. 02 Predict

    The cerebellum forecasts the sensory consequences of every motor command. Latent world models roll a policy's future forward before it acts, and a few probing motions identify the dynamics of an unfamiliar object before the first real attempt.

    Intercepting the Future · Wiggle and Go!

  3. 03 Correct

    Prediction error drives behavior. Imagined rollouts are checked against the manifold of demonstrated behavior, so drift is caught and recovered from before it happens physically.

    FAR · Ariadne's Clew (in progress)

  4. 04 Consolidate

    The brain replays experience and strengthens only the connections that matter. Vision-language-action policies adapt online through experience replay and parameter-efficient fine-tuning of the layers that count, which sharpens what the next cycle perceives.

    ExpReS-VLA · Not All Layers Need Tuning · Hierarchical Skill Retrieval

Technical Lead, Applied Research at Stealth, and Visiting Scholar at the Carnegie Mellon Robotics Institute.

  • ICRA 2026 Best Paper finalist
  • CMU Robotics Institute Outstanding Graduate Research Assistant Award and Best Robotics Technology Student, 2025
  • CVPR, ICRA, CoRL, RSS, IROS, ISMAR and a granted US patent

Publications and patents

Work in progress

  1. 2027
    THING: Translating Haptics Into Native-Token Grasping

    S.N. Syed, A. Jakobsson, P. Sachdev, J. Ichnowski

    In preparation for Robotics: Science and Systems (RSS 2027)

  2. 2027
    Preview coming soon
    Ariadne's Clew: Predictive Temporal Alignment of Imagined Rollouts to Human Demonstrations

    S.N. Syed, A. Jakobsson, P. Sachdev, J. Ichnowski

    In preparation for the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2027)

Under review

  1. 2027
    Not All Layers Need Tuning: Diagnosing and Directing Adaptation in Vision-Language-Action Models

    S.N. Syed, A. Jakobsson, P. Sachdev, J. Ichnowski

    Submitted to the IEEE International Conference on Robotics and Automation (ICRA 2027)

  2. 2026
    Hierarchical Skill Retrieval for Data-Efficient Adaptation of Vision-Language-Action Models

    H. Hao, S.N. Syed, J. Schneider, J. Ichnowski

    Under review at IEEE Robotics and Automation Letters (RA-L)

Conferences

  1. 2026
    Intercepting the Future: Latent-Space Predictive World Model for Dynamic VLA Manipulation

    S.N. Syed, A. Jakobsson, H. Hao, J. Ichnowski

    Conference on Robot Learning (CoRL 2026)

  2. 2026
    FAR: Failure-Aware Retry for Test-Time Recovery and Continual Policy Improvement

    H. Hao, S.N. Syed, J. Ichnowski, J. Schneider

    Conference on Robot Learning (CoRL 2026)

  3. 2026
    Wiggle and Go! System Identification for Zero-Shot Dynamic Rope Manipulation

    A. Jakobsson, A. Mahajan, K. Pullalarevu, K. Suresh, Y. Mao, B.P. Duisterhof, Y. Yao, S.N. Syed, J. Ichnowski

    Conference on Robot Learning (CoRL 2026)

  4. 2026
    ExpReS-VLA overview. An RGB frame and the prompt place the white mug in the off-white bowl go into a frozen OpenVLA vision backbone with SigLIP and DINOv2, combined with retrieval, a prioritizer, LlamaIndex, a compressed replay buffer and the policy, trained and deployed on an NVIDIA RTX 5090, producing actions for a Franka 7-DOF arm
    ExpReS-VLA: Specializing Vision-Language-Action Models Through Experience Replay and Retrieval

    S.N. Syed, Y. Ahuja, A. Jakobsson, J. Ichnowski

    IEEE International Conference on Robotics and Automation (ICRA 2026) Best Paper finalist

  5. 2021
    Learning by Aligning Videos in Time

    S. Haresh, S. Kumar, H. Coskun, S.N. Syed, A. Konin, M.Z. Zia, Q.H. Tran

    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2021)

  6. 2020
    RetroActivity: Rapidly Deployable Live Task Guidance Experiences

    A. Konin, S.N. Syed, S. Siddiqui, S. Kumar, Q.H. Tran, M.Z. Zia

    IEEE International Symposium on Mixed and Augmented Reality (ISMAR 2020), demonstration Best Demo Award

  7. 2020
    The assistive mobile robot, its AutoCAD design beside the built robot, a differential-drive base with ultrasonic sensors and an RPLidar, a controller box, an assistive box, speakers, and an LCD face with localization and surveillance cameras on top
    Assistive Mobile Robot for Industrial and Academic Applications

    S.R.N. Jafri, A. Ahmed, A. Azam, U.B. Ihsan, S.N. Syed, R. Uddin

    International Bhurban Conference on Applied Sciences and Technology (IBCAST 2020)

  8. 2016
    ICONICS2016
    Greenhouse Monitoring and Automation System

    S.N. Syed, A.Z. Shaikh

    International Conference on Innovations in Computer Science and Software Engineering (ICONICS 2016)

Journals

  1. 2018
    Mehran University Research Journal2018
    A Novel Hybrid Biometric Electronic Voting System: Integrating Fingerprint and Face Recognition

    S.N. Syed, A.Z. Shaikh, S. Naqvi

    Mehran University Research Journal of Engineering and Technology, 37(1)

  2. 2017
    IJAAS2017
    Remote Monitoring System for Diesel Generators at Cellular Base Stations

    S.N. Syed, A.Z. Shaikh, S. Naqvi, T. Altaf

    International Journal of Advanced and Applied Sciences, 4(8)

Patents

  1. 2025
    Bump Detection for Facilitating a Defined Vehicle Stop Behavior

    J. Shah, S. Jha, D.R. Pamnani, A. Pulapaka, P. Oza, T. Eto, T. Aguiar, L. Pedersen, Z. Guter, P.N. Shah, S.S. Undavalli, S. Muthusamy, S.N. Syed

    US patent application with Nissan North America, Inc. · Docket NNA2-5544-A · Filed July 2025 Pending

  2. 2025
    Automated Mapping for Localization in a Controlled Environment

    S. Jha, P. Oza, L. Pedersen, A. Pulapaka, D.R. Pamnani, J. Shah, T. Eto, Z. Guter, P.N. Shah, S.S. Undavalli, S.N. Syed, S. Muthusamy

    US patent application with Nissan North America, Inc. · Docket NNA2-5546-A · Filed July 2025 Pending

  3. 2025
    Autonomous Navigation in Controlled Environments

    S. Jha, P. Oza, L. Pedersen, A. Pulapaka, D.R. Pamnani, J. Shah, Z. Guter, P.N. Shah, S.S. Undavalli, S.N. Syed, S. Muthusamy

    US patent application with Nissan North America, Inc. · Docket NNA2-5547-A · Filed July 2025 Pending

  4. 2022
    System and Method for Correlating Video Frames in a Computing Environment

    Q.H. Tran, M.Z. Zia, A. Konin, S. Haresh, S. Kumar, S.N. Syed

    US Patent 11,368,756 Granted

Preprints

  1. 2026
    Keypoints found by XFeat, ALIKED and SuperPoint, left to right, on a motion-blurred room and a cluttered industrial scene. The counts differ sharply by extractor and scene.
    CHAMELEON-SLAM: Adaptive Feature Selection and Uncertainty-Aware Matching for Robust Monocular Visual SLAM

    S.N. Syed

    TechRxiv preprint

  2. 2025
    Input frames and the reconstructed views from the method, side by side, for a ballroom and an indoor corridor.
    Joint Learning of Depth, Pose, and Local Radiance Field for Large Scale Monocular 3D Reconstruction

    S.N. Syed, Y. Hu, Y. Yao

    arXiv preprint arXiv:2512.18237

  3. 2025
    Visuo-Acoustic Hand Pose and Contact Estimation

    Y. Mao, U. Yoo, Y. Yao, S.N. Syed, L. Bondi, J. Francis, J. Oh, J. Ichnowski

    arXiv preprint arXiv:2508.00852

  4. 2025
    Preview coming soon
    SuperPoint-SLAM3: Augmenting ORB-SLAM3 with Deep Features, Adaptive NMS, and Learning-Based Loop Closure

    S.N. Syed, I. Roongta, K. Ravie, G. Nageswar

    arXiv preprint arXiv:2506.13089

Experience

  1. Jun 2025 – now

    Technical Lead, Applied Research

  2. Jun 2025 – now

    Visiting Scholar, Momentum Lab

  3. Jun 2024 – Aug 2024

    Research Intern, Perception and Autonomy

  4. Aug 2023 – Apr 2025

    Research Assistant

  5. Jun 2022 – Jul 2023

    Senior Robotics Engineer (Contractor)

  6. Jul 2019 – May 2021

    Contractor (SBIR)

  7. Nov 2018 – Jul 2023

    Founding Engineer and Lead Research Engineer

Education

  1. 2023 – 2025

    M.S. Robotic Systems Development · GPA 4.00

    Outstanding Graduate Research Assistant Award and Best Robotics Technology Student, Class of 2025.

  2. 2014 – 2018

    B.E. Electronic Engineering, with distinction

    Ranked 9th of 143. Undergraduate researcher in the Autonomous Systems Lab.

News

  1. Nov 2026Talk

    Invited talk at Harvard Medical School, hosted by Pierre Dupont, on world action models and their use in model predictive control for cardiac surgery.

  2. Oct 2026Award

    Named an Outstanding Reviewer at CoRL 2026.

  3. Sep 2026Paper

    Not All Layers Need Tuning submitted to ICRA 2027.

  4. Sep 2026Paper

    Three papers accepted to CoRL 2026: Intercepting the Future, FAR and Wiggle and Go!.

  5. Aug 2026Recognition

    Awarded a Certificate of Appreciation by the Executive Office of the President, White House Office of Science and Technology Policy, for serving as an expert judge in the first Presidential AI Challenge.

  6. Aug 2026Paper

    Hierarchical Skill Retrieval under review at IEEE Robotics and Automation Letters.

  7. Jul 2026Award

    Best Reviewer Award at the ICML 2026 Second Workshop on Agents in the Wild (AIWILD).

  8. Jun 2026Service

    Expert judge for the first Presidential AI Challenge at the state, regional and national levels, with the national round at the White House.

Earlier news
  1. Jan 2026Paper

    ExpReS-VLA accepted to ICRA 2026, later named a Best Paper finalist.

  2. Nov 2025Award

    Outstanding Graduate Research Assistant Award and Outstanding Graduate Teaching Assistant Award (Honorable Mention), CMU Robotics Institute, 2025.

  3. Jun 2025Career

    Joined Stealth as Technical Lead, Applied Research, and the CMU Momentum Lab as a Visiting Scholar.

  4. Apr 2025Award

    Best Robotics Technology Student Award, CMU Robotics Institute, Class of 2025.

  5. Jun 2024Career

    Started as an autonomy and perception intern at Nissan North America.

  6. Aug 2023Career

    Started the M.S. in Robotic Systems Development at the Robotics Institute, Carnegie Mellon University.

  7. Jul 2022Recognition

    Retrocausal named to the CB Insights Advanced Manufacturing 50, the top 50 of more than 6,000 advanced manufacturing startups.

  8. Jun 2022Patent

    Granted US Patent 11,368,756, System and Method for Correlating Video Frames in a Computing Environment.

  9. May 2021Award

    Retrocausal won the Judges' Award at the 2021 Vision Tank start-up competition.

  10. Mar 2021Paper

    Learning by Aligning Videos in Time accepted to CVPR 2021.

  11. Nov 2020Award

    ISMAR 2020 Best Demo Award for RetroActivity: Rapidly Deployable Live Task Guidance Experiences.

  12. Aug 2020Grant

    Funded by a NASA JPL SBIR grant to build a few-shot learning computational model with the Exploration Medical Capability (ExMC) laboratories for crewed Mars missions.

  13. Nov 2018Career

    Joined Retrocausal AI as a Founding Engineer.

Contact

The fastest way to reach me is email.

shahramnajam2022@gmail.com LinkedIn Google Scholar GitHub

Presidential AI Challenge certificate awarded to Shahram Najam Syed for contributions to the 2026 Presidential AI Challenge, from the Executive Office of the President, Office of Science and Technology Policy.
ICML Certificate of Achievement recognizing Shahram Najam Syed with the Best Reviewer Award for outstanding contribution to the Second Workshop on Agents in the Wild (AIWILD), awarded July 11, 2026, signed by the workshop chairs.
Page 1 of the U.S. Department of Energy, Office of Science, Acknowledgement of Service for the 2026 Presidential Artificial Intelligence Challenge, dated April 13, 2026, confirming Shahram Najam Syed's service as an expert judge at the state, regional and national tiers. Page 2 of the letter, describing the state-level judging across nine states, regional judging for the East Central Region on April 4 and 13, 2026, and the invitation to judge at the national tier. Page 3 of the letter, confirming the judging duties were fulfilled with diligence and professionalism, signed by Jan Tyler, Program Manager, Presidential AI Challenge, U.S. Department of Energy, Office of Science.
2025 Robotics Institute Outstanding Graduate Research Assistant Award presented to Shahram Syed, November 2025, signed by Srinivasa G. Narasimhan, Interim Director of the Carnegie Mellon University Robotics Institute, and Martial Hebert, Dean of the School of Computer Science. 2025 Robotics Institute Outstanding Graduate Teaching Assistant Award, Honorable Mention, presented to Shahram Syed, November 2025, signed by Srinivasa G. Narasimhan and Martial Hebert. Carnegie Mellon University Robotics Institute certificate of achievement awarded to Shahram Najam Syed, Best Robotics Technology Student, MRSD Class of 2025, signed by John M. Dolan, MRSD Program Director, and Sarah Conte, MRSD Program Manager.
ISMAR 2020 Best Demo Award, 19th International Symposium on Mixed and Augmented Reality, presented to Andrey Konin, Shahram Najam Syed, Shakeeb Siddiqui, Sateesh Kumar, Quoc-Huy Tran and M. Zeeshan Zia of Retrocausal Inc. for the demo RetroActivity: Rapidly Deployable Live Task Guidance Experiences, signed by the four general co-chairs, November 9 to 13, 2020, virtual.