Varun Gandhi
I'm an MS Computer Science student at UMass Amherst, interested in AI research and engineering, with a focus on LLM post-training. I have experience across both research and industry. Outside of research, I write notes on things I'm learning, inspired by the idea of learning in public.
- 2026
This summer, I'm interning as a Software Engineer at Motional. - 2026
I was an AI Research Extern at Adobe (Jan–May 2026) through the UMass CS 698DS Industry Mentorship Practicum in AI, working on LLM post-training, agents, and retrieval-augmented generation for document Q&A — mentored by Zichao (Jack) Wang at Adobe and Jaewook (Jake) Lee at UMass. This work became GRASP. - 2025
I'm pursuing my MS in Computer Science at UMass Amherst. In Fall 2025, I was the instructor for four sections of undergraduate CS First Year Seminars. - 2025
I interned as a Machine Learning Engineer with the Vision-Language Models team at Sarvam AI, where I contributed to their efforts in building India's sovereign language model. - 2024 Awarded the Bay State Fellowship by Manning CICS — a merit-based award granted to around ten students each year, with a teaching assistantship and full tuition waiver.
News
- Jul 2026 New preprint: "GRASP: GRanularity-Aware Search Policy for Agentic RAG", from my internship work with Adobe Research.
- Jun 2026 Our paper "Hierarchical Experimentalist Agents" (HExA) was accepted to the Third Reinforcement Learning Beyond Rewards Workshop at the Reinforcement Learning Conference (RLC).
- Summer 2026 Interning at Motional as a Software Engineer Intern.
- Sep 2025 Instructor for four sections of undergraduate CS First Year Seminars at UMass this fall.
- Summer 2025 Interned as a Machine Learning Engineer with the Vision-Language Models team at Sarvam AI.
- Jan 2025 Began my MS in Computer Science at UMass Amherst.
- Oct 2024 Awarded the Bay State Fellowship by Manning CICS.
Research
GRASP: GRanularity-Aware Search Policy for Agentic RAG
arXiv preprint, 2026
GRASP uses reinforcement learning to train an agentic RAG policy that decides when to retrieve, which tool to use (semantic search, keyword search, or paragraph reading), and at what granularity, collecting sentence-level evidence as needed. The learned policy develops interpretable skimming and scanning behavior and improves retrieval recall and QA accuracy on multi-hop benchmarks.
Hierarchical Experimentalist Agents
arXiv preprint, 2026
HExA is a training-free framework that lets LLM agents improve through active experimentation — learning reusable skills and integrating evidence to solve novel tasks without external supervision — evaluated on Interphyre, a benchmark built on the PHYRE 2D physics environment.
Chain-of-Code Collapse: Reasoning Failures in LLMs via Adversarial Prompting in Code Generation
arXiv preprint, 2025
We show that LLMs are fragile under semantically faithful but adversarially structured prompt variations in code generation — performance swings on surface-level formatting changes rather than genuine reasoning ability.