AI Researcher
Physical Intelligence · Computer Vision · Robot Learning
Building physical AI systems that can perceive, learn, reason, and act in the physical world (a.k.a. robots that can be absolutely delightful and obviously helpful).
Currently exploring
SO-101 Robot Learning
Investigating imitation learning and vision-language-action policies for real-world manipulation.
Featured research
Unified ML Pipeline for Lunar Terrain Age Determination
An end-to-end ML pipeline that classifies and clusters 1.3M+ lunar impact craters into age-based clusters.
AIM 2020 Challenge on Rendering Realistic Bokeh
Team submission to the AIM 2020 challenge on synthesizing realistic shallow-depth-of-field (bokeh) effects from all-in-focus images.
Low-Light Face Detection
Exploring robust object detection under challenging illumination conditions using transformer-based computer vision models.
Writing
View all writing →Understanding World Models
Coming soonWhat are world models, what do they actually learn, and why might they matter for physical intelligence?
Learning for Physical Intelligence
Coming soonNotes on robot learning, reinforcement learning, vision-language-action models and embodied intelligence.
About
I am an AI researcher interested in physical intelligence — how machines can learn to perceive, reason about, and act in the physical world.
My background is in applied electronics and instrumentation engineering, software engineering, and computer vision research — a combination that maps directly onto what physical AI demands: sensing and interfacing with real hardware, perceiving the world through vision, and engineering it all into systems that hold up outside a simulator.
I am currently exploring robot learning, world models, reinforcement learning, and generalization in embodied systems.
Long-term research interest
My long-term research interest is in building general-purpose intelligent robots that can learn, adapt, and operate in the physical world, while helping make capable robots accessible and useful to everyone.
I am particularly interested in the learning, reasoning, and adaptation mechanisms that could enable robots to become more capable, flexible, and autonomous across tasks, environments, and embodiments.
Research interests
Background
Sept 2026 — Present
Career transition into physical AI · Self-directed
Full-time, focused work on robot learning fundamentals and a small number of rigorous, well-documented hardware projects.
June 2025 — Present
Research Assistant · University of Kerala, Geology Department (under Dr. Sajin Kumar K S)
Designed and led an end-to-end ML pipeline classifying 1.3M+ lunar impact craters by age — supervised classification at 95.6% accuracy, followed by ensemble clustering into age-based groups. First-author paper accepted at ICAITA 2026, but not published — unable to attend the conference in person to present it; currently preparing to resubmit to a new venue.
Aug 2021 — Dec 2023
Applications Engineer · Oracle India
Built and shipped web and mobile UI features for enterprise data-integration workflows; authored APIs and test automation to improve release reliability.
Jun 2019 — Jul 2021
Undergraduate Researcher, Computer Vision Lab · College of Engineering, Trivandrum (under Dr. Jiji C V)
Built a DETR-based object detection model for low-light face detection; competed in international CV challenges at ECCV 2020 and CVPR 2021.