Anirban

The rapid advancement of Large Language Models (LLMs) has fundamentally reshaped intelligent systems, enabling natural language reasoning, code generation, and complex decision support. More recently, the emergence of agentic AI has shifted the research focus from single-model interactions toward autonomous systems capable of planning, memory management, tool utilization, multi-step reasoning, and collaborative problem solving. While these capabilities demonstrate significant promise, translating research prototypes into enterprise-scale deployments remains an open systems engineering and AI research challenge. This talk presents a research-oriented perspective on the design of enterprise-ready agentic systems. We examine the architectural foundations of autonomous AI agents, including planning and reasoning strategies, Retrieval-Augmented Generation (RAG), structured and long-term memory, tool invocation, workflow orchestration, and multi-agent collaboration. Particular emphasis is placed on emerging design patterns that enable scalable, modular, and resilient agentic applications capable of operating in dynamic, high-risk environments. Beyond system architecture, the presentation explores current research challenges in evaluating autonomous AI systems. Topics include benchmarking agent reasoning, measuring task completion and robustness, mitigating hallucinations, ensuring explainability, quantifying uncertainty, and incorporating human oversight into autonomous decision-making. The discussion also covers safety, governance, observability, and compliance considerations that are essential for deploying agentic AI in domains such as fraud detection, cybersecurity, financial services, healthcare, and enterprise operations. Drawing from recent advances in generative AI and practical experience developing production-scale AI systems, the talk bridges the gap between cutting-edge research and industrial deployment. It concludes by highlighting open research opportunities in self-improving agents, adaptive memory architectures, agent communication protocols, lifelong learning, and trustworthy multi-agent ecosystems. Attendees will gain a comprehensive understanding of the scientific principles, engineering trade-offs, and research directions that will shape the next generation of autonomous enterprise AI systems.

Biography:

Anirban Majumder is an Applied Scientist II at Amazon and an AI/ML leader with more than 20 years of experience designing and deploying machine learning systems at enterprise scale. His work spans generative AI, large language models, fraud detection, risk modeling, intelligent agents, causal inference, and production machine learning. Throughout his career at Amazon and leading technology consulting organizations, Anirban has led the development of AI solutions that have delivered tens of millions of dollars in measurable business impact. His recent work focuses on agentic AI systems, LLM evaluation, responsible AI, safety benchmarking, and scalable multi-agent workflows for enterprise applications. Anirban is a Senior Member of IEEE, Member of ACM and has authored multiple peer-reviewed publications in artificial intelligence, fraud detection, AI agents, and healthcare AI, including research publications at ACM SIGKDD and EMNLP. He is also a co-inventor on granted patents spanning AI-driven materials science and digital health technologies. In addition to his research, he serves as a technical reviewer for books and mentors scientists and engineers on production AI systems and responsible deployment of generative AI. His interests include building trustworthy autonomous AI systems that combine advanced reasoning capabilities with robust governance, safety, and measurable business outcomes.

Email: