Undergrad in Mathematics & Computing
- Computer Vision
TopoReformer (Accepted at the 5th Workshop on AI for CyberSecurity @AAAI 2026)
Mitigating Adversarial Attacks Using Topological Purification in OCR Models
- LLMs
Stress testing eval awareness in LLMs
Training model organisms that (evade activation-based eval-awareness probes) U (non-verbalised) while preserving eval-aware behavior.
Benchmarking SLMs on quantization and RLHF effects to measure bias and fairness across standard evaluation datasets.
Simulation of LLM agent collusion under Cournot competition strategies
- Community & Open Source Contributions
Chrome extension that lets you share your open-source conversations




