Context, Retrieval & Compromised Agents

Two talks on the unglamorous end of running AI agents: getting the right context into them, and what happens when one gets taken over. António Morais and Akbar Gumbira, both machine learning engineers at Elastic, cover building better context with visual intelligence and semantic retrieval - how to feed agents the right material and serve the models behind them reliably. Shade Obiti, a cybersecurity analyst at Accenture, follows with a live demo of a compromised AI agent that keeps coming back after the obvious security measures are in place. Hosted by Manchester AI at Accenture's Circle Square office. Registration is gated: RSVP yes, complete the organiser's form, and bring photo ID on the night. Why go: retrieval quality and agent security are where most production AI work actually goes wrong, and both talks come from people shipping it. Best for ML and platform engineers.