I am an independent AI safety researcher and final-year B.Tech Computer Science & Engineering (AI) undergraduate at Mar Baselios College of Engineering and Technology in Thiruvananthapuram, India. My research sits at the intersection of sequential statistical auditing and multimodal robustness—building open-source evaluation tools that detect capability drift, shortcut reliance, and adversarial fragility in Vision-Language Models (VLMs) without wasting compute on fixed-horizon sweeps.

Previously, I built and open-sourced VLM-TRACE, a modular evaluation pipeline for testing multimodal provenance and adversarial robustness under white-box PGD+EOT perturbations and diffusion edits. I am the lead author of "Drift-Aware Anytime-Valid Monitoring of Training Integrity: The Innovation E-Detector and Calibrated Detection from the Training Loss" (accepted at IEEE ICICAIEECE-2027), and I have contributed open-source pull requests (PR #3) to UBC TEA Lab's vlm-probe repository for counterfactual VLM shortcut evaluation alongside publishing reproducible negative-results stress tests on Zenodo.

Alongside research, I serve as Chairperson of the IET MBCET Student Chapter and represented my institution as a panel speaker at the IET International Festival of Volunteering in London in September 2026. I am currently building E-Audit, an open-source sequential evaluation wrapper and UK AISI inspect_ai plugin, ahead of applying to Fall 2027 research graduate programs in machine learning and AI safety.