The GÉANT School of Software Engineering (SSE) returns from 20-22 October 2026 with a new three-day, face-to-face workshop exploring Advanced Knowledge & Agentic Systems with LLMs.
Supporting Software Development Across the Community
The School of Software Engineering continues to evolve alongside the needs of the GÉANT community, helping software development teams strengthen and broaden their software engineering skills. The programme has also widened its audience and many SSE training events no longer require specific prior knowledge, making them accessible to a broader range of participants interested in developing their technical expertise.
Advanced Knowledge & Agentic Systems with LLMs
This SSE workshop builds on last year’s Buzz-Free Programming with LLMs. This time, the focus moves beyond first experiments to the practices that make LLM-based systems dependable in production.
Over three days, participants will explore how to take retrieval beyond naive setups, helping answers remain accurate and relevant at scale. The sessions will look at how to organise and govern knowledge so systems can remain trusted, up to date, and able to reason over how things connect rather than simply matching text. Participants will also learn how to build knowledge layers that improve over time instead of starting from scratch on every query.
The second half of the workshop focuses on agentic systems. Participants will explore how to design systems that can plan, use tools, and break down complex problems into steps. Sessions will also examine when autonomous agents are the right choice and how to keep them reliable while maintaining human oversight.
Throughout the workshop, the focus remains practical and buzz-free, helping participants understand when a technique adds value and when a simpler approach is a better choice. 70% of the workshop will be hands-on exercises in a Python environment, with the remainder delivered as lectures. Participants should bring a laptop with a working IDE and some experience building LLM-based systems.
Advanced Retrieval and Knowledge Base Design
The first day focuses on where naive RAG breaks in production, and how to harden retrieval (pre- and post-retrieval optimisation, modular pipelines). Participants will also explore metadata as a first-class context layer, how to govern a knowledge base for reliability, and how to build a modular, metadata-aware retrieval pipeline.
Reasoning over Knowledge: Agentic, Graph and Self-Maintaining RAG
The second day expands on these foundations by introducing approaches such as Agentic RAG, Graph RAG, how to build and curate knowledge, and extend the pipeline with agentic and graph-based retrieval.
Agentic Systems: Frameworks, Workflows and Deep Agents
The final day turns to agentic systems. Sessions cover topics including tool use, planning, reflection, multi-agent systems, workflows vs. autonomous agent, and deep agents. Attendees will then apply these concepts by developing an agent capable of breaking down and solving a domain problem.
