I am Rahul Ghosh, an Applied Scientist at the Amazon AWS GenAI Innovation Center in Santa Clara, California. My research focuses on building agentic AI systems, multi-modal foundation models, and LLM evaluation frameworks.

My Work and Research

My current work at AWS spans several areas:

• Agentic AI Systems: Developing frameworks for visual compliance verification using multi-modal LLMs (CompAgent, CVPR’26) and agentic retrieval-augmented generation for telecom specifications (TelcoAI, AACL-IJCNLP’25).
• LLM Evaluation: Research on probabilistic frameworks for evaluating LLM self-consistency and logical reasoning.
• Foundation Models: Previously developed foundational representation models for Amazon Music and spatio-temporal foundation models for earth observation tasks.

My research interests include:

• Core: Foundation Models, Sequence Modeling, Computer Vision, Multi-Modal Modeling, Representation Learning
• Methods: Self-supervised Learning, Few-shot & Zero-shot Learning, Meta Learning, Variational Inference
• Applications: Uncertainty Quantification, Causal Inference, Scientific AI, Earth Observation

My Background

I received my Ph.D. in Computer Science from the University of Minnesota (2018–2023), advised by Prof. Vipin Kumar, where my dissertation focused on “Entity-Aware Knowledge-Guided Machine Learning for Scientific Knowledge Discovery.” Prior to that, I received my B.Tech in Electronics & Electrical Engineering from IIT Guwahati (2012–2016).

Before my PhD, I worked at Samsung R&D Institute in Delhi as a Research Engineer, focusing on predictive analytics for IoT sensor data and NLP for customer service applications.

I have published 50+ papers in top venues including CVPR, KDD, AAAI, ICDM, IJCAI, SDM, Nature, and Water Resources Research, and have received multiple best paper awards (IEEE ICDM 2021 & 2023, ACM SIGSPATIAL 2022, SIAM SDM 2022).