About Me
I am a computer scientist in the Mathematics and Computer Science Division at Argonne National Laboratory, where I develop next-generation foundation models and intelligent AI systems to accelerate scientific discovery at scale. My research lies at the intersection of agentic AI, foundation models for science, multimodal reasoning, trustworthy machine learning, and high-performance computing, to enable AI systems that can reason over complex scientific data, collaborate autonomously, and operate reliably in scientific environments.
I lead and contribute interdisciplinary research across multiple scientific domains, including infrastructure resilience, intelligent transportation systems, HPC networks, hazard adaptation, and scientific visualization through several DOE projects.
Prior to joining Argonne, I served as a Senior Data Scientist at General Electric, where I applied advanced machine learning to large-scale industrial systems. I earned my Ph.D. in Computer Science from the Indian Institute of Technology (IIT) Kharagpur, India.
My current research interests
- Agentic systems: Our research investigates how autonomous AI agents can plan, reason, and coordinate in complex, dynamic environments, with a focus on designing scalable orchestration and workflow intelligence that enable reliable and efficient multi-agent decision-making.
- Multimodal Reasoning for Scientific Discovery:
We study how AI systems reason across text, images, time series, and spatiotemporal data, focusing on representation learning, cross-modal alignment, and evaluation of multimodal scientific understanding.
- Foundation Models training: We study how foundation models can be trained and adapted for scientific domains while preserving generalization, reasoning, and scalability, with an emphasis on data-efficient and domain-aware learning for reliable scientific use.
- Reliable AI systems:
We study methods for ensuring that AI systems are robust, interpretable, and dependable in high-stakes environments. We focus on uncertainty quantification, hallucination detection, failure prediction, and risk-aware reasoning to support safe and predictable deployment in real-world scientific applications.