Hi! I'm Bhavesh, a Dual Degree (B.Tech + M.Tech) student in Computer Science & Engineering at IIT Delhi, graduating in May 2026. I work broadly on natural language processing and large language models — spanning diffusion language models, dialogue systems, and the security of reasoning models.

Most recently, I'm a Research Intern at Microsoft Research, developing conversational agents for medical diagnosis. I'm also an AI Security Intern at the National University of Singapore (KISP Lab) with Prof. Prateek Saxena, studying compute-amplification attacks on LLMs.

My M.Tech thesis, advised by Prof. Parag Singla, develops Hierarchical Masked Diffusion Language Models for controllable, non-autoregressive generation. Earlier, I built probabilistic dialogue systems for medical diagnosis with Prof. Mausam, and worked on post-training infrastructure for long-context agentic LLMs at Neosigma.

Beyond research, I've worked as a Quantitative Researcher at Ebullient Securities — building and live-testing alpha signals and C++ trading strategies — and on compiler tooling and formal verification at CompilerAI Labs under Prof. Sorav Bansal.