Critical Software at Sword AI Summit 2026: Intelligence for a Deterministic World
Critical Software's Sérgio Pereira takes the stage at Sword AI Summit 2026 in Porto to share how ontologies, evals, and evidence trails make AI agents trustworthy in safety-critical engineering.
AI agents are everywhere except where the stakes are highest. Safety-critical engineering doesn't have room for 'usually right’. So how do you put AI to work in a world built on determinism, traceability, and zero tolerance for ambiguity?
That's the question Sérgio Pereira, Cluster Technology Director at Critical Software, will take on at the Sword AI Summit 2026, taking place October 3rd at the Alfândega Congress Centre in Porto.
The Talk: Intelligence for a Deterministic World
Sérgio's session sets out three foundations for making AI agents work in domains where failure isn't an option:
Ontologies that encode engineering judgment, so the system reasons in the domain's own terms — not a generic approximation of it.
Evals that turn trust into numbers, replacing gut feel with something you can measure and defend.
Evidence trails that make every decision traceable for audit, because in safety-critical work, an answer without a paper trail isn't an answer. The through-line: the binding constraint on AI in these environments isn't the model. It's the knowledge.
Why it Matters
This is the exact tension Critical Software navigates every day — building software that keeps trains running, aircraft flying, and medical devices working, while exploring how AI can genuinely earn a place in that world rather than just being bolted onto it. Sérgio's talk brings that frontline experience to a stage shared with speakers from NVIDIA, Google DeepMind, OpenAI, and Sword itself.
Register for Sword AI Summit 2026 — we'll see you there.