AI-Driven Optical Transport Networks: Architectures, Applications, and Intelligent Automation

SC463 - AI-Driven Optical Transport Networks: Architectures, Applications, and Intelligent Automation

07 Mar 2027
13:30 - 17:30

Short Course Level

Intermediate

Short Course Description

This short course examines how agentic AI can transform the operation and management of optical transport networks. It introduces the core architectural building blocks, shows how they can be combined into operations-ready solutions, and connects them to practical network use cases. Software-defined networking (SDN) and network programmability are introduced briefly as enabling functionality for accessing network data, exposing control capabilities, and integrating AI agents with operational systems. The short course will discuss used cases and examples covering network planning, service provisioning, predictive maintenance and root cause analysis, and network optimization. Participants will learn how basic building blocks of AI Agentic can be integrated to simplify operations, accelerate troubleshooting, and enable closed-loop optimization across optical and packet layers.

The course is structured in two parts: Part 1 introduces the key building blocks for agentic AI in optical networks, while Part 2 translates these concepts into practical use cases, applications, demonstrations, and prototypes.

Part 1 — Building Blocks for Agentic AI in Optical Networks
Part 1 develops a practical architecture for agentic AI in optical networking. It explains the role and interaction of the following building blocks:

• Large language models (LLMs) — for natural-language interaction, reasoning, intent interpretation, and workflow support.
• Retrieval-augmented generation (RAG) and GraphRAG knowledge bases — for grounding responses and actions in network documentation, topology, inventory, telemetry, operational procedures, and historical data.
• Model Context Protocol (MCP) — for connecting AI agents to tools, data sources, network controllers, and operational applications through standardized interfaces.
• Agent-to-Agent (A2A) communication — for coordinating specialized agents across domains, layers, and operational responsibilities.
• Intelligence functions — for integrating local machine-learning models, optimization algorithms, rule engines, and complex workflows that invoke external tools, simulators, controllers, or assurance systems.

The course discusses how these components support tool use, knowledge grounding, multi-agent collaboration, policy enforcement, guardrails, observability, and human-in-the-loop control. It also highlights the requirements for trustworthy, secure, and explainable automation in carrier-grade optical networks.

Part 2 — Use Cases, Deployment Options, Demos, and Prototypes
Part 2 explores how agentic AI can be applied to network operations through representative use cases, deployment models, demonstrations, and prototypes. Topics include:

• Conversational access to network knowledge, topology, inventory, alarms, and performance data
• Intent translation, service planning, and what-if analysis
• Fault detection, diagnosis, localization, and guided remediation
• Predictive maintenance, health monitoring, and capacity headroom assessment
• Cross-layer and multi-domain coordination using cooperating agents
• Human-in-the-loop workflows and ChatOps for controlled operational automation
• Local ML inference, optimization functions, and compound workflows involving external tools, simulations, and operational systems
• Deployment options ranging from assistant-style copilots to policy-controlled multi-agent systems

Practical demonstrations and prototypes illustrate how agents can retrieve network knowledge, invoke operational tools, collaborate with other agents, and execute governed workflows. The course concludes with deployment considerations, architectural trade-offs, limitations, and a roadmap toward autonomous—but accountable—optical network operations.

Keywords: optical networks, agentic AI, LLMs, RAG, GraphRAG, MCP, A2A, network automation, network intelligence, SDN, network programmability, multi-layer networks, multi-domain networks, operations automation.

Short Course Benefits

By the end of this course, participants will understand how agentic AI building blocks can be combined with programmable network infrastructures and operational workflows to support more intelligent, explainable, and accountable optical network automation.

• Understand the key building blocks of agentic AI for optical networks, including LLMs, RAG and GraphRAG knowledge bases, MCP-based tool integration, agent-to-agent coordination, and local intelligence functions.
• Explain how SDN, network programmability, open data models, telemetry, and controller interfaces provide the operational foundation for AI-assisted and agent-driven network automation.
• Discuss how agentic AI can support network planning, service provisioning, what-if analysis, fault diagnosis, predictive maintenance, root cause analysis, and network optimization.
• Understand the role of ChatOps, LLM agents, and human-in-the-loop workflows in simplifying service provisioning, troubleshooting, network change management, and operational decision support.
• Assess the benefits and operational risks of agentic AI, including explainability, safety, governance, policy enforcement, observability, and compliance logging for AI-supported actions.
• Evaluate cross-layer and multi-domain automation scenarios across optical and packet layers using open interfaces, intent frameworks, policy constraints, and specialized collaborating agents.
• Identify practical deployment options ranging from assistant-style copilots to policy-controlled multi-agent systems, supported by demonstrations and prototypes.

Short Course Audience

This course is designed for network and system architects, engineers, researchers, and technical managers working on packet-optical network automation, multilayer optimization, and AI-driven operations. Attendees should have a basic understanding of AI and network operation concepts and be interested in deepening their knowledge of architectures, applications, and implementation.

Instructors

• Achim Autenrieth

  ADVA Optical Networking SE, Germany

  About the Instructor

  Achim Autenrieth is Senior Director Advanced Technology at Adtran Networks SE, where he is leading the research activities on networking technologies including intelligent network automation and SDN control of 5G/6G and disaggregated optical transport networks, machine learning, and planning and evaluation of multilayer optical networks. Achim is member of IEEE and VDE/ITG, he authored or co- authored more than 120 reviewed and invited scientific publications and he is technical program committee member of OFC (2018-2021, 2024-2026), ECOC (2011-2017), ONDM, DRCN and RNDM.

• Jörg-Peter Elbers

  ADVA Optical Networking SE, Germany

  About the Instructor

  Jörg-Peter Elbers is VP of Advanced Technology, Standards & IPR at Atran, responsible for global technology strategy, applied research, standardization, and intellectual property. He has over 25 years of experience in the telecommunications industry, with previous technical leadership roles at ADVA, Ericsson, Marconi, and Siemens. Jörg holds a Dr.-Ing. (PhD) degree in Electrical Engineering from the Technical University of Dortmund.