2025: Published 'mtRAG: A Multi-Turn Conversational Benchmark for Evaluating Retrieval-Augmented Generation Systems' (ACL 2025)
2025: Proposed 'Instruction-Following with Dynamic Attention Steering'
2025: Introduced 'Training with Pseudo Code for Instruction Following'
2025: Released dataset with 2500+ invocable API endpoints derived from NL2SQL databases
2024: Proposed 'Reducing the Scope of Language Models'
2025: Published work on tools for tracking adherence to AI model behavioral use clauses
2024: Recognized as one of IBM’s top global technical contributors
IBM Technical Award recipient (2015, 2019, 2025)
2018: Named to MIT Technology Review’s Innovators Under 35 – India
2020: British Council Alumni Award for Professional Achievement – India
2017: Inaugural member of the ACM Future of Computing Academy
2016: Romberg Scholar at the 4th Heidelberg Laureate Forum (HLF)
Research Experience
Leading research on conversational AI and large language models at IBM Research AI
Developing methods to enhance LLM performance in agentic Retrieval-Augmented Generation (RAG) and API tasks with multi-hop reasoning and grounding
Researching training- and inference-time approaches to improve instruction-following capabilities of LLMs
Chaired the IEEE-SA P2840 working group and BigScience Model Governance Working Group, leading to the first large LLM released under Behavioral-use Licensing
Collaborated with AAAI to enable RAIL licensing for code/models associated with conference papers