I am a senior research engineer working on language research at Google Deepmind (formerly Google Research - Brain Team). My current research focus is on improving the generation and alignment capabilities of large language models, specifically RLHF and preference optimization for post training of Gemini. Before joining Google, I did my research masters in Language Technologies (MLT) from Language Technologies Institute, School of Computer Science at Carnegie Mellon University, co-advised by Alan W Black, Yulia Tsvetkov and Alex Rudnicky.
Before I joined CMU, I worked in Samsung Research Institute, Bangalore, India where I was working in the Voice Intelligence Team on dialogue systems and chat-bots. Before that, I spent a wonderful semester working with Partha Talukdar at Machine and Language Learning (MALL) Lab in the Indian Institute of Science, Bangalore. I graduated with a B.Eng. in Computer Science from Birla Institute of Technology and Science, Pilani, India. In my Bachelor's thesis, I worked on incorporating external knowledge in distantly supervised neural relation extraction methods as part of iNELL (which is based on NELL).
News
We released Gemini: a family of highly capable multimodal models.
Our work on Statistical rejection sampling improves preference optimization was accepted in The Twelfth International Conference on Learning Representations (ICLR) 2024.
Our work on Calibrating Sequence likelihood Improves Conditional Language Generation was accepted in The Eleventh International Conference on Learning Representations (ICLR) 2023.
Our work on Unsupervised Keyphrase Extraction via Interpretable Neural Networks was accepted at the 2023 Conference on European Association for Computational Linguistics.
I joined Google Research to continue my reserach on language generation with the Brain Team.
Our work on Improving Broad-Coverage Medical Entity Linking with Semantic Type Prediction and Large-Scale Datasets was accepted at the Journal of Biomedical Informatics.
Our work on DialoGraph: Incorporating Interpretable Strategy-Graph Networks into Negotiation Dialogues. was accepted at the International Conference on Learning Representations (ICLR) 2021.
Our work on ResPer: Computationally Modelling Resisting Strategies in Persuasive Conversations was accepted at the European Association for Computational Linguistics (EACL) 2021.
Our work on MedType: Improving Medical Entity Linking with Semantic Type Prediction was accepted at the AMIA 2021 Virtual Informatics Summit.
Our work on Keeping Up Appearances: Computational Modeling of Face Acts in Persuasion Oriented Discussions was accepted at the 2020 Conference on Empirical Methods in Natural Language Processing.
Our work on Incorporating Multi-Level Features for Multi-Granular Propaganda Span Identification was accepted at the 14th International Workshop on Semantic Evaluation, 2020.
Our team, Tartan, reached the semi-finals of the 3rd Alexa Dialogue Challenge, 2019.
Our work on Analyzing the Extent of Misinformation in Cancer Related Tweets was accepted for poster presentation at ICWSM 2020.
Our work on A Multistream Vector Representation for use in Natural Dialog was accepted for poster presentation at LREC 2020.
Gave a talk at the 2019 LTI Student Research Symposium. Won a honourable mention award.
Our work on Improving Distanty Supervised Neural Relation Extraction using Side Information was accepted for oral presentation at EMNLP 2018.