Scholar
Lakshya A Agrawal
Google Scholar ID: WwW9ij4AAAAJ
University of California, Berkeley
Citations & Impact
Citations
179
H-index
5
i10-index
4
Publications
11
Co-authors
32
list available
Publications
Resume (English only)
Academic Achievements
- Led the development of 'GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning', introducing a novel approach for optimizing LLM-based pipelines through interpretable, language-driven reflection. Also led the development of dspy.GRPO, the first GRPO pipeline for tuning modular agents, including complex compound AI Systems. Led the work on 'Monitor-Guided Decoding of Code LMs with Static Analysis of Repository Context', proposing Monitor Guided Decoding (MGD), which was accepted at NeurIPS '23 and won first place in Microsoft Global Hackathon.
Research Experience
- Worked as an AI4Code Research Fellow at Microsoft Research, focusing on improving the code generation capabilities of Large Language Models (LLMs) and exploring how generative AI can automate software engineering tasks. Current research focuses on enhancing the quality and correctness of code generated by LLMs, aiming to improve their reliability for software engineering and reasoning tasks.
Education
- A Computer Science and Artificial Intelligence PhD student at UC Berkeley, advised by Prof. Matei Zaharia and Prof. Dan Klein.
Background
- Research interests span Artificial Intelligence, Software Engineering, and Programming Languages. Prior to joining UC Berkeley, he was an AI4Code Research Fellow at Microsoft Research.
