Pivotal Research - Under Tyler Tracy, Redwood ResearchResearch Fellow
London, UK, Feb 2026 – Present
Building control protocols and evaluations for coding agents in realistic software engineering settings.
Amazon AWSSWE Intern
Bangalore, India, Jan 2026 – Feb 2026
Worked as a part of AWS Opensearch Team
AI Safety CampResearch Fellow
Remote, Jan 2026 - Present
Working on 2 projects: Mechanistic Interpretability and AI Control
Project 1: Understanding Temporal reasoning in large language models
Project 2: Building new Monitoring protocols
Algoverse ResearchAI Safety Fellow
Remote, August 2025 – Present
Exploring cooperative AI and delegation in principal–agent settings.
Built formal models for value and capability misalignment, designed constrained synthetic environments to study trust and autonomy trade-offs, and ran grid-world experiments to quantify when delegation remains rational under misalignment.
Algoverse ResearchMachine Learning Researcher
Remote, June 2025 – Present
Working on mechanistic interpretability and emergent misalignment.
Identified a shared 2D misalignment subspace across Qwen models, studied layer-wise LoRA weight geometry, and developed suppression and gating methods that reduce misaligned behaviors at inference time.
Tech4Good Lab, UC Santa CruzMachine Learning Research Intern
Santa Cruz, CA, July – September 2025
Worked with Prof. David Lee on multi-agent simulations for large community events.
Adapted 1,000-agent architectures to real interview data and built behaviour-conditioning pipelines that outperformed demographic and persona baselines.
PI School of AIAI Fellow
Rome (Hybrid), Italy, June – July 2025
Designed an explainable anomaly detection system for a manufacturing partner.
Tested a wide range of resampling and modelling strategies and helped identify process variables responsible for scrap reduction and real cost savings.
Colorado State UniversityMachine Learning Research Intern
Fort Collins, CO, May – July 2025
Extended LM-based sandbox evaluators to study safety risks in autonomous agents.
Built over twenty adversarial test scenarios for tool-use environments and improved automatic safety scoring pipelines using GPT-4.
Infysec SolutionsSoftware Engineering Intern
Remote (Chennai), February – April 2025
Helped build a production cybersecurity training platform now used by security engineers and pentesters.
Led frontend development for challenge workflows and shipped dozens of hands-on labs backed by containerized and isolated environments.
EDUCATION
Amrita Vishwa VidyapeethamBachelor of Technology in Computer Science and Engineering
Coimbatore, India · Sep 2022 – June 2026
Relevant coursework
CS:
Operating Systems, Algorithms and Data Structures, Computer Networks, Database Management Systems, Machine Learning, Deep Learning, Distributed Systems, Secure Coding, Compiler Design, Principles of Programming Languages
Math:
Linear Algebra, Probability and Statistics, Discrete Mathematics, Number Theory, Optimization Techniques