I am a fourth year PhD student at McGill University, Graduate Research Assistant at Mila - Quebec AI Institute, and Visiting Researcher at ServiceNow Research. I am advised by Siva Reddy. I am curious about the interplay of knowledge and language in language models. My current research focuses on improving text representations, specifically for retrieval-augmented generation (RAG) applications. I am also interested in making conversational models more trustworthy and reliable to ensure equitable distribution of information.
My research journey started at Data Analytics and Intelligence Research (DAIR) Group, IIT Delhi where I was advised by Mausam. Before that I was a software developer at Samsung Research Institute Bangalore. I hold a Bachelor of Technology in Mathematics and Computer Science from IIT Guwahati (2013-17).
When I’m not behind a laptop, I like to spend my time outdoors biking, playing tennis, ice-skating, or skiing.
Updates:
- April 2024: New paper! LLM2Vec: Large Language Models Are Secretly Powerful Text Encoders. We propose a simple efficient recipe to convert any decoder-only LLM into a universal text encoder. Check out the paper, blog post, tutorial, Twitter/X announcement, and HuggingFace models.
- March 2024: I participated in a panel discussion on Human-centered Metrics for Search-Oriented Conversational Artificial Intelligence (SCAI).
- February 2024: Our paper Evaluating Correctness and Faithfulness of Instruction-Following Models for Question Answering has been accepted at TACL, an ACL-sponsored journal published by MIT Press. It will be presented at ACL 2024, to be held in Bangkok, Thailand. Check out the paper and code for more details.
- December 2023: In Winter 2024, I will be a TAing COMP-345: From Natural Language to Data Science taught by Siva Reddy and Morgan Sonderegger. This is my third time TAing this course.
- August 2023: New paper! Evaluating Correctness and Faithfulness of Instruction-Following Models for Question Answering. We advocate for a new evaluation paradigm for retrieval-augmented question answering and propose simple token-overlap metrics that correlate highly with human judgments.
- April 2023: I gave a tutorial at IVADO Labs on Question answering: from the document to the excerpt.
- February 2023: I am excited to join ServiceNow Research again as a visiting researcher. I will be working with Dzmitry Bahdanau on evaluation of QA models and LLM-based sentence representations.
- December 2022: In Winter 2023, I will be a TAing COMP-345: From Natural Language to Data Science taught by Siva Reddy and Tim O’Donnell. This is my second time TAing this course.
- October 2022: I have been awarded Graduate Excellence fellowship by McGill University.
- June 2022: I presented TopiOCQA at Mila Partner Symposium, an annual event held to foster collaboration between researchers at Mila and its industry partners.
- Janurary 2022: In Winter 2022, I will be a TAing COMP-424: Artificial Intelligence taught by Jackie Chi Kit Cheung.
- December 2021: Our paper TopiOCQA: Open-domain Conversational Question Answering with Topic Switching has been accepted at TACL, an ACL-sponsored journal published by MIT Press. It will be presented at ACL 2022, to be held in Dublin, Ireland. Check out the project website for more details.
- October 2021: I participated in a panel discussion on Question Answering Evaluation at 6th Workshop on Search-Oriented Conversational AI (SCAI 2021), held online. The recording is available on YouTube.
- August 2021: I will be a part-time visiting researcher at ServiceNow Research (previously ElementAI), where I will be working with Harm de Vries on open-domain conversational question answering.
- August 2021: Our paper Regex Queries over Incomplete Knowledge Bases has been accepted at AKBC 2021. The conference will be held virtually.
- July 2021: In Fall 2021, I will be a TAing COMP-596: From Natural Language to Data Science taught by Siva Reddy and Tim O’Donnell.
Website template and design stolen with ❤️ from Desh.