Avik Dutta

Hi! I am Avik Dutta, a pre-doctoral Research Fellow with the PROSE team at Microsoft. I am particularly interested in Natural Language Processing, LLM4Code, Information Retrieval and AI4SE (Software Engineering). I am also keen on exploring areas involving efficient code generation through in-context learning, controlling decoding strategies through neuro-symbolic approaches, and helping language model's performance through retrieval.

I completed my undergraduate (B.Tech) in Electronics and Electrical Communication Engineering from Indian Institute of Technology Kharagpur, India. Additionally, I received a Minor degree in Computer Science and Engineering, and a Micro Specialization in Artificial Intelligence and Applications. During this time, I was honored to work with Prof. Plaban Kumar Bhowmick on projects involving Graph Neural Networks and Reinforcement Learning. Following that, I got the opportunity to work as an Undergraduate Student Researcher in CNeRG, under the guidance of Prof. Animesh Mukherjee. I worked on novel problems which included Named Entity Recognition, domain specific QA, instruction tuning LLMs, etc.

I worked briefly as a data analyst in Piramal Finance before I joined Microsoft as a Research Fellow. At Microsoft, I am working on problems related to NL2Code, under the guidance of Dr. Vu Le and Dr. Sumit Gulwani. I am also fortunate to be advised by Dr. Ashish Tiwari, for the work on context inference in spreadsheets, and Reasoning/Planning strategies for processing NL queries in Excel Copilot. Additionally, I am also exploring problems on solving Advanced Data Analysis tasks using Multi-Agentic frameworks in Excel Copilot, under the supervision of Dr. Arjun Radhakrishna and Dr. Gustavo Soares.