Please note: This seminar will take place in DC 1304.
Anwar Walid, PhD
Modern networked systems operate under continual change, yet most control mechanisms encode objectives at design time, limiting adaptability under partial observability and delayed feedback. This talk examines a shift toward runtime context-conditioned control. I begin with control-theoretic and reinforcement learning approaches for multipath transport (MPTCP), showing how learning can operate inside real protocol stacks under strict stability constraints. I then describe how AI foundation models can be embedded as decision components, enabling adaptation based on context and intent without redesigning underlying controller logic. Examples from low-latency congestion control (L4S) and service-level resource control illustrate how such mechanisms can be embedded in real systems while preserving safety and deployability. I conclude by outlining directions for scalable, adaptive infrastructure at the intersection of networking, control, and AI.
Bio: Anwar Walid’s research spans networked and distributed systems, control, optimization, and artificial intelligence. His work focuses on decision-making under uncertainty in large-scale infrastructure, including transport protocols and adaptive resource management. He previously led Network Intelligence and Distributed Systems Research at Nokia Bell Labs, where his work influenced deployed systems and international standards. He is an ACM Fellow, IEEE Fellow, and AAIS Fellow. He has served on numerous National Science Foundation review panels and related workshops and is an Adjunct Professor at Columbia University. He received his Ph.D. from Columbia University and his B.S. and M.S. degrees from New York University.