Department of Computational and Data Sciences
Department Seminar
Speaker : Saravan Rajmohan, Chetan Bansal and Victor Ruehle, M365 Research Group, Microsoft
Title : A Full-stack perspective to AI Efficiency: Experiences from M365 Research at Microsoft
Date & Time: December 09th, 2025 (Tuesday), 14:00 PM
Venue : # 102, CDS Seminar Hall
ABSTRACT
The explosive growth of Generative AI has resulted in significant efficiency challenges and rising operational costs. Addressing these challenges requires a holistic, cross-stack optimization approach that spans models, AI systems, infrastructure, and hardware. M365 Research is an Applied Research group at Microsoft advancing the state-of-the-art in AI and Systems research, organized around Cloud reliability, AI efficiency, and next-gen architecture for Productivity Agents. In this talk, we will explore the key bottlenecks and identify opportunities for improving efficiency across all layers of the technology stack. We further showcase illustrative projects to reduced cost and enhance scalability of generative AI solutions.
BIO: Saravan Rajmohan leads the M365 Research group at Microsoft. He is an industry-leading expert and visionary leader in AI systems, infrastructure, generative AI, and large language model (LLM) systems, he holds numerous top-tier publications and patents. Saravan is focused on building world-class generative AI systems that are efficient by design, grounded, trustworthy, private, and most importantly, personalizable and adaptable across domains. His global team of over 40 researchers is at the forefront of AI and systems innovation.
Chetan Bansal is a Senior Principal Research Manager at Microsoft. He is passionate about designing data-driven tools and techniques for step function improvement in Cloud Reliability, Efficiency and Developer Productivity. At Microsoft, he has built from scratch and is leading an interdisciplinary applied research team (AI, NLP, Systems, Software Engineering) of ~20 researchers and research engineers. His work has directly contributed to significant COGS savings, reliability gains for 100+ services, and productivity gains for 1000+ engineers at Microsoft. He has published 50+ papers in top international conferences and has 25+ patents. His research has been recognized with awards at FMCAD, SoCC and ICSE conferences.
Victor Ruehle is a Senior Principal Research Manager at Microsoft, where he leads an applied research team driving innovations to improve efficiency of generative AI scenarios. His work bridges research and product, with an emphasis on full-stack optimizations across models, AI systems, infrastructure, and hardware. Victor earned his PhD in physics from the Max-Planck-Institute for Polymer Research in Mainz, Germany, in 2010. Before joining Microsoft, he worked as a physical property specialist in the oil and gas industry and as a postdoctoral researcher at the Department of Chemistry, University of Cambridge.
Host Faculty: Yogesh Simmhan, CDS
ALL ARE WELCOME
