Pritish Chakraborty
Hi! I am an MS+PhD student in the CSE department at IIT Bombay, advised by Prof. Abir De and Prof. Soumen Chakrabarti. My research is in neural information retrieval and efficient search, with a focus on settings where standard top-K retrieval is insufficient. This includes multi-hop and compositional queries, retrieval over structured objects such as graphs, and set-valued retrieval objectives where the retrieved items must collectively satisfy an information need.
A recurring theme in my work is to enable indexing and retrieval for complex information needs. This includes the development of learned representations that enable both classical inverted indexing and dense indexing, and approximation algorithms that support richer notions of relevance, while remaining efficient at scale.
I graduated from the MS by Research program at IIT Bombay in June 2024, advised by Prof. Abir De, and subsequently continued into the PhD program. Before joining IIT Bombay, I worked as a full-stack software engineer at Instahyre, a recruitment startup building candidate-job matching products. I was promoted to Head of Engineering for my contributions and enjoyed the experience of building and maintaining production systems.
news
| Apr 30, 2026 | Two papers - “(Position) Neural Approximation Is Rarely Justified for Hard Combinatorial Problems” and “Multi-Way Representational Alignment” - accepted to ICML 2026. |
|---|---|
| Jan 26, 2026 | Our work - A Dense Subset Index for Collective Query Coverage (DISCo) - has been accepted at ICLR 2026! |
| Dec 9, 2025 | Indradyumna Roy and I presented our work on a novel dense subset index for collective information retrieval at Snowflake Inc. |
| Oct 11, 2025 | Received the NeurIPS 2025 Scholar Award, which covers registration and accommodation during the conference. |
| Sep 18, 2025 | Our paper on Contextual Tokenization for Graph Inverted Indices (CoRGII) has been accepted at NeurIPS 2025 as a poster! Looking forward to attending in December. |