Half-day Tutorial: Optimizing Low-Latency Video Streaming: AI-Assisted Codec-Network Coordination

The tutorial will take place Online

• Farzad Tashtarian (Alpen-Adria-Universität - AAU)
• Zili Meng (Hong Kong University of Science and Technology - HKUST)
• Abdelhak Bentaleb (Concordia University)
• Mahdi Dolati (Sharif University of Technology)

Session 1 | Introduction: Network and Application Layers in Video Streaming
08:45 — 09:15	Presentation I | Zili Meng Near-Bufferless Ultra-Reliable Low-Latency Video Streaming
09:20 — 09:50	Presentation II | Abdelhak Bentaleb App Meets Transport: Rethinking Video Streaming over QUIC
Session 2 | Video Streaming Optimization: AI- and Network-Assisted Approaches
09:55 — 10:25	Presentation I | Farzad Tashtarian End-to-End Adaptive Video Streaming Optimization
10:30 — 11:00	Break
11:00 — 11:30	Presentation II | Mahdi Dolati Programmable Networks for Video Streaming
Session 3 | From Test benches to Practice: Hands-On Demonstration
11:35 — 12:00	Ultra-low latency communication and video standardization activity | Sergey Ikonin
12:00 — 12:35	Getting your hands dirty with AI-assisted video codecs | Jiayang Xu

This tutorial focuses on the emerging need for ultra-low-latency video streaming and how AI-assisted coordination between codecs and network infrastructure can significantly improve performance. Traditional end-to-end streaming pipelines are often disjointed, leading to inefficiencies under tight latency constraints. We present a cross-layer approach that leverages AI for real-time encoding parameter adaptation, network-aware bitrate selection, and joint optimization across codec behavior and transport protocols. The tutorial examines the integration of AI models with programmable network architectures (e.g., SDN, P4) and modern transport technologies such as QUIC and Media over QUIC (MoQ) to minimize startup delay, stall events, and encoding overhead. Practical use cases and experimental insights illustrate how aligning codec dynamics with real-time network conditions enhances both QoE and system efficiency. Designed for both researchers and engineers, this session provides a foundation for developing next-generation intelligent video delivery systems capable of sustaining low-latency performance in dynamic environments.

Low-latency video streaming is critical for emerging applications such as live events, telemedicine, industrial teleoperation, and immersive media. However, traditional streaming architectures often treat codecs and network protocols as separate layers, resulting in inefficiencies under dynamic conditions. This tutorial addresses the need for tightly integrated codec-network coordination, where AI plays a central role in enabling real-time adaptation, semantic-aware traffic prioritization, and end-to-end optimization. By combining advances in AI-driven encoding, network programmability (e.g., SDN, P4), and modern transport protocols like QUIC and MoQ, we aim to equip participants with the knowledge and tools to design the next generation of low-latency streaming systems.

• Session (1) - Introduction: Network and Application Layers in Video Streaming (60 min)
• Session (2) - Video Streaming Optimization: AI- and Network-Assisted Approaches (60 min)
• Session (3) - From Testbenches to Practice: Hands-On Demonstration (60 min)

This tutorial is designed for researchers and practitioners seeking to understand and contribute to the development of next-generation intelligent media delivery systems.

Farzad Tashtarian is a Postdoctoral Researcher at the Department of Information Technology, Alpen-Adria-Universität Klagenfurt, Austria. He received his Ph.D. in Computer Engineering from Ferdowsi University of Mashhad, Iran. His research focuses on end-to-end video streaming systems, network-assisted multimedia delivery, adaptive bitrate (ABR) algorithms, and energy-aware streaming strategies. He has a particular interest in integrating artificial intelligence and machine learning into multimedia systems to improve Quality of Experience (QoE), resource efficiency, and sustainability. Farzad has published extensively on video streaming optimization, cloud/edge resource management, and AI-based delivery mechanisms. He is currently leading the end-to-end aspects of multimedia delivery in the Christian Doppler Laboratory ATHENA project and is actively involved in organizing workshops and mentoring students. His recent work emphasizes cross-layer innovations, including network optimization, in-network computation, generative AI, and client-side intelligence for both live and VoD streaming applications. More details can be found at: https://tashtarian.net/

Zili Meng is an assistant professor in the Department of Electronic and Computer Engineering at Hong Kong University of Science and Technology (HKUST). He received his B.Eng. (Hons) and Ph.D. (Hons) from Tsinghua University. His current research interest focuses on ultra-low latency video streaming from all layers, resulting in 9 papers in SIGCOMM and NSDI in recent years. He is the recipient of Doctoral Dissertation Awards from ACM China and Chinese Institute of Electronics, the Gold Medal of SIGCOMM'18 and SIGCOMM'24 Student Research Competition, and two best paper awards.

Abdelhak Bentaleb is an assistant professor in the Department of Computer Science and Software Engineering (CSSE) at Concordia University, where he is also the founder and director of the IN2GM Lab (Intelligent Networking and Networked Multimedia Systems Group). His research lies at the intersection of networking systems, multimedia, systems, and applied AI, with a strong focus on designing and building large-scale, data-driven, sematic-based end-to-end solutions for networked multimedia systems. His work aims to address real-world challenges by leveraging cutting-edge technologies to optimize performance, efficiency, and user experience. Dr. Bentaleb earned his Ph.D. from the National University of Singapore (NUS), where his dissertation focused on enabling optimizations in HTTP Adaptive Streaming (HAS) for video delivery. This groundbreaking research earned him two highly prestigious awards: the SIGMM Award for Outstanding Ph.D. Thesis and the DASH Industry Forum Best Ph.D. Dissertation Award. Following his Ph.D., he continued his research at NUS as a Postdoctoral Research Fellow for three years, further advancing his expertise in multimedia streaming and adaptive video delivery systems. Dr. Bentaleb is a co-inventor on three patents, and his research has led to the publication of over 60 papers in top-tier conferences and journals. Throughout his career, he has been recognized with several prestigious awards for his contributions to the field. His ongoing projects reflect his ambition to bridge the gap between theory and practice in largescale networked multimedia systems through applied AI and data-driven methodologies. More details can be found at: https://users.encs.concordia.ca/~abentale/

Mahdi Dolati is an Assistant Professor at Sharif University of Technology, Tehran, Iran. He received his Ph.D. in Computer Engineering from the University of Tehran. His research focuses on the design and measurement of programmable and softwarized networks to support emerging network-oriented applications such as distributed machine learning and video streaming. He is particularly interested in addressing the complex challenge of allocating scarce network resources through theoretically grounded approaches. Mahdi has published on topics including consistent packet forwarding in Software-Defined Networks (SDNs), resource allocation for Virtual Networks, machine learning-based network control, and caching in programmable networks. He also has experience in developing network simulators for modern virtualized and containerized environments. Currently, he leads a team of graduate students conducting research on next generation networking systems. His recent work emphasizes in-network machine learning, distributed ML acceleration, and simulating deep learning clusters.

N/A