AI observability as the engine of production-ready AI

Samuel Jaja takes the Leeds Sharp stage on Thursday 21 May to deliver a talk grounded in a real multi-agent financial compliance system. The case study is a five-agent pipeline handling OFAC sanctions screening, AML detection and regulatory rules enforcement, with every agent decision captured in an immutable Langfuse audit trail showing token usage, latency, cost and data lineage in real time. This is not a theoretical overview — it is a walkthrough of what observability actually looks like inside a production AI pipeline.

The talk covers three architecturally distinct approaches to building the same system: raw SDK with custom parallel orchestration via asyncio; the OpenAI Agents SDK for rapid prototyping; and MCP (Model Context Protocol) for exposing tools across team and organisational boundaries. Each trades differently, and the talk is explicit about when to reach for which. Attendees will leave with a mental model for when multi-agent architecture is genuinely necessary, why observability is the feedback loop that compresses build-debug-ship cycles, and how Azure OpenAI changes the calculus when data sovereignty and auditability are non-negotiable in regulated industries such as finance, energy and construction.

The format is a single talk with pizza and drinks from 6pm, talk starting at 7pm. Leeds Sharp is a .NET user group that meets regularly in Leeds city centre. The audience skews towards .NET engineers and software architects, but the observability and multi-agent content is relevant to anyone building production AI systems regardless of stack. Sponsors include Fruition Group (.NET recruitment) and JetBrains. Some attendees head to the Editors Draft pub afterwards at 8pm.

Why attend

Samuel Jaja builds production AI systems at Capgemini UK across regulated industries. He founded PyData Hull, holds an MSc in AI and Data Science with Distinction from the University of Hull, and spoke at DDD North 2026. The material is drawn from real systems, not demos.

If you are deciding between asyncio orchestration, the OpenAI Agents SDK and MCP for a current project, this talk gives you a concrete basis for that decision. The Langfuse audit trail angle is directly applicable if you work in any regulated sector where auditability is a hard requirement. Pizza and drinks are provided, and there is a pub afterwards.

Agenda

• 18:00 — Doors open, pizza and drinks
• 19:00 — Talk: From Black Box to Glass Box: Systems Thinking & Observability as the Engine of Production-Ready AI — Samuel Jaja
• 20:00 — Pub (Editors Draft, vote to be taken on the night)

Speakers

• Samuel Jaja — Generative AI Engineer at Capgemini UK; founder of PyData Hull and Atlasync AI Ltd; MSc AI and Data Science with Distinction, University of Hull; speaker at DDD North 2026.