Building Reliable AI Agents: From Context to Evals
Details
Building Reliable AI Agents: From Context to Evals
The inaugural London Agentic AI Meetup brings together researchers, practitioners, and builders exploring the frontier of Agentic AI systems. Our first session dives deep into what it takes to make AI agents powerful and reliable from the way we engineer context, to how we evaluate and optimize outputs when tasks are fuzzy, subjective, or open-ended.
This evening will explore the latest thinking around:
- Context Engineering – shaping inputs for dependable agent performance.
- Evaluation & Optimization with DSPy – closing the loop on prompts and tasks.
- LLM-as-a-Judge – training models to evaluate generative quality.
- Data Infrastructure Foundations – semantic layers, ontologies, and knowledge graphs as critical scaffolding for reliable AI applications.
Speakers
🎤 Mike Taylor (LinkedIn)
Talk: The Evaluator–Optimizer Pattern in DSPy
Mike will demonstrate how to train an LLM-as-a-Judge and then use it to optimize fuzzy generative tasks — where evaluations are informal, subjective, and difficult to measure. This approach unlocks new ways of making AI agents more dependable in creative and ambiguous domains.
🎤 Leonardo Ubbiali (LinkedIn)
Talk: Context Engineering for Reliable AI Agents
Leonardo will share insights on the importance of the semantic layer/ontology - structuring the data context - How you feed the data context into the agentic workflow (from prompt engineering to context engineering)
Sponsor
This meetup is proudly sponsored by the authors of the book:
📖 Prompt Engineering for Generative AI: Future-Proof Your Career
Why Attend?
Whether you’re building AI agents, experimenting with DSPy, or simply curious about the next wave of reliable AI systems, this meetup is for you. Expect:
✅ Practical insights into evaluator–optimizer loops.
✅ Best practices in context engineering.
✅ Guidance on semantic data layers for reliability.
✅ Networking with London’s growing agentic AI community.
Agenda (6:00 – 9:00 pm)
- 6:00 – 6:30 – Arrival, networking, refreshments, Networking
- 6:45 – 7:30 – Talk 1: Mike Taylor
- 7:30 – 7:50 – Talk 2: Leonardo Ubbiali
- 7:50 – 8:00 – Panel Q&A
- 8:00 – 9:00 – Networking
***
🔗 Join us for London Agentic AI Meetup #1 and be part of shaping how we build reliable AI agents from context to evals.
Venue
Thanks to AutogenAI for providing venue.
Building Reliable AI Agents: From Context to Evals