The 2025 edition hosted four workshops and two tutorials on 26 October, the third day of the symposium. The workshop and tutorial day was curated by Elena Vasquez (Workshops Chair) and Lucas Moreau (Tutorials Chair), with proceedings and notes archived alongside the main programme.

Workshops

W1 — AI-Native Data Centers

Organizers: Karim Hassan (ETH Zürich), Jonas Fischer (TU München), Wei Chen (Peking University).

A workshop on learned congestion control, in-network telemetry, and data-plane ML inside hyperscale data centers. The programme included contributed talks on RDMA congestion control, learned load balancing, and SmartNIC-accelerated inference, plus a panel on the operational realities of deploying ML inside the network fabric.

W2 — Federated Learning at the Wireless Edge

Organizers: Sara Bertrand (Sorbonne Université), Ravi Ramanathan (IIT Madras), Yusra Khalid (NUTECH).

A workshop on the practical challenges of federated learning across bandwidth-constrained, intermittently connected wireless links. Topics included gradient quantization under link budgets, client selection under mobility, and the interaction between federated optimization and MAC-layer scheduling.

W3 — Benchmarking and Reproducibility for Networked AI

Organizers: Helena Pereira (INESC-ID), Mei Han (Zhejiang University), Lukas Steiner (TU Berlin).

A workshop on datasets, benchmarks, and reproducibility tooling for the networking/AI intersection. Featured a live artifact-evaluation session, an open dataset showcase, and a town-hall discussion on community benchmarks. The workshop proceedings fed directly into the Best Artifact award deliberations.

W4 — LLM Agents for Network Operations

Organizers: Ricardo Almeida (Universidade de Lisboa), Hyunwoo Ahn (KAIST), Pavlos Vassileiou (FORTH).

The newest workshop of the 2025 edition, dedicated to LLM-based agents for configuration synthesis, incident response, and operator tooling. Featured contributed talks on retrieval-augmented generation over operational runbooks, tool-using agents for closed-loop remediation, and evaluation methodology for agentic network operations.

Tutorials

T1 — Building a Programmable Data Plane with P4 and eBPF

Organizer: Adriana Kowalski (University of Warsaw).

A full-day tutorial covering the practical construction of programmable data planes using P4 and eBPF. The tutorial included hands-on exercises on match-action pipelines, parser state machines, and porting simple ML inference primitives onto a software switch and a SmartNIC.

T2 — Practical Federated Learning over Real Networks

Organizer: Carolina Meneses (Universidad de los Andes).

A half-day tutorial on the practical engineering of federated learning systems deployed over real, bandwidth-constrained networks. Covered gradient compression, secure aggregation, client scheduling under link heterogeneity, and the use of open-source frameworks (Flower, PySyft) over simulated WAN topologies.