Alok Raj

I am a researcher focused on Robotic Perception, Manipulation, and Vision-Language Models. My work spans vision-based models for autonomous agents, including visual active search, task-oriented grasping, and multi-agent navigation, aiming to create adaptive robots that integrate advanced perception, learning, and control. I wish to advance robot intelligence through vision transformers, reinforcement learning, and 3D vision techniques that enable autonomous decision-making in dynamic environments.

I am currently a CS undergraduate at IIT Dhanbad, India. Most recently, I have been a research intern at the Multi-Agent Robotic Motion (MARMoT) Lab, NUS under Prof. Guillaume A. Sartoretti, working on embodied vision-language models for visual search with test-time adaptation methods and manipulation policy mobilization. I have also worked at the Center of Intelligent Robotics, IIIT Allahabad under Prof. G.C. Nandi and Andrew Melnik, developing the GRIM framework for generatively conditioned task-oriented grasping, and at Samsung R&D Institute India, building low-compute speaker verification systems. Previously, I interned at Clutterbot Technologies, where I improved robot vision models through self-training, knowledge distillation, and curriculum learning.