Noushad Sojib

Robot Learning Engineer building robust policies from imperfect human demonstrations. Focused on imitation learning, VLA models, and diffusion policies for real-world robotic systems.

M.Sc. in Computer Science from the University of New Hampshire (Cognitive Assistive Robotics Lab). B.Sc. from SUST, where I founded a robotics club and built humanoid robots from scratch.

Currently seeking roles in robot learning and real-world robotic systems.

Email  /  GitHub  /  Google Scholar  /  LinkedIn

Research

How can robots reliably learn from imperfect human teachers? My research tackles this challenge head-on— developing algorithms that enable robots to extract safe, generalizable behaviors from noisy or erroneous demonstrations, bringing robot learning one step closer to real-world deployment.

Self Supervised Detection of Incorrect Human Demonstrations: A Path Toward Safe Imitation Learning by Robots in the Wild

Real-world robot learning breaks down when humans give imperfect demonstrations—yet perfect data is rarely available. We present BED (Behavior Cloning for Error Detection), a self-supervised framework that automatically identifies and filters out incorrect human demonstrations, enabling robots to learn safe policies from messy, real-world data. Validated in RoboSuite simulation and on a physical Sawyer robot arm.

Self-Supervised Visual Motor Skills via Neural Radiance Fields

We introduce a NeRF-powered visual imitation learning architecture that learns motor policies in a fully self-supervised manner—no human-labeled training data required. By combining neural radiance fields with key-point correspondence, the system builds rich 3D scene understanding and generalizes robot manipulation skills from raw visual observations.

Robot From Scratch

Long before studying robot learning formally, I was already building robots from the ground up. Inspired by ASIMO but with no local robotics community, I founded RoboSUST—a student robotics club—and led a team that built humanoid robots from scratch with minimal resources. This hands-on foundation in mechanics, embedded systems, and control gives me a rare end-to-end perspective on robotics.

Ribo Robot

Ribo was showcased at multiple national events across Bangladesh, drawing wide coverage in print and broadcast media and inspiring a new generation of robotics enthusiasts.

Lee: A biped walking robot

See the baby steps

Kiddo

Hardware Design

Building robots from scratch taught me that good software needs good hardware. Along the way, I designed several embedded devices—a few of which landed in peer-reviewed venues and are now in active use on research robots.

3Wheel Mouse

Islam, Md Touhidul, et al. “Wheeler: A three-wheeled input device for usable, efficient, and versatile non-visual interaction.” ACM UIST 2024. Paper & Video

Charging Dock

Live demonstrated at IROS 2023. An extended version is actively used with the Hello Stretch robot. See example

Lowcost Braille Display

Sojib, Noushad, and M. Zafar Iqbal. “Single cell bangla braille book reader for visually impaired people.” IEEE ICBSLP 2018. Paper

Design and source code from Jon Barron's website