May meetup: Building valuable products in a probabilistic world

Leeds Data Science Meetup's May 2026 session brings two practitioners to The Elbow Room on Call Lane to talk about what actually changes when you start building with probabilistic systems.

John Carney, Principal Consultant at PDFTA and co-founder of PyDataMCR, opens with a talk titled Building Valuable Deterministic Products in a Probabilistic World. His argument: almost everything in software has historically been deterministic, and LLMs break that assumption in ways that require different thinking. He'll share concrete techniques and principles for building products that hold up when the underlying model doesn't give you the same answer twice.

Obinna Iheanachor, AI/Data Engineer at Rotork and founder of Wisabi Analytics, follows with a case study that has a number worth paying attention to: he reduced the cost of processing 4,700+ engineering drawings from £8,000 to under £40 by combining deterministic methods with LLMs rather than relying on a cloud-only approach. The talk covers when to call a model, how to manage cost versus accuracy trade-offs, and what breaks at scale in a real UK engineering company.

The format is a standard Leeds Data Science evening: food from 18:00 on a first-come, first-served basis, welcome at 18:45, then two 35-minute talks running back to back until 20:00. The event is free to attend.

Leeds Data Science Meetup is organised in partnership with Jumping Rivers, a Leeds-based data science consultancy. The May session is hosted by Parallax, a digital agency based in Leeds. If you want to speak at a future event, contact lds@jumpingrivers.com.

The audience tends to be working data scientists, ML engineers and analysts. Both talks are pitched at practitioners who have moved past the tutorial stage and are thinking about production systems.

Why attend

Obinna's cost reduction case study — £8,000 to £40 on a real engineering workload — is the kind of number that makes a talk worth attending on its own. The specifics around when to call a model and how to manage accuracy trade-offs are directly applicable if you're building anything with LLMs in production.

John Carney brings a broad view from working across startups and large organisations, and as chair of the PyData London Conference he's well-connected in the UK data community. Free entry, food provided, central Leeds venue. Worth the Tuesday evening.

Agenda

• 18:00–18:45 — Refreshments (food on a first-come, first-served basis)
• 18:45–18:50 — Welcome
• 18:50–19:25 — John Carney (PDFTA): Building Valuable Deterministic Products in a Probabilistic World
• 19:25–20:00 — Obinna Iheanachor (Rotork): From £8,000 to £40: What I Learned Shipping AI in a UK Engineering Company

Speakers

• John Carney — Principal Consultant at PDFTA; co-founder of PyDataMCR; chair of PyData London Conference and the PyData Strategic Committee; co-founder of the Field of Play conference. Has delivered £10s of millions in value across data projects spanning startups to large corporates.
• Obinna Iheanachor — AI/Data Engineer at Rotork, building production-grade AI and data systems for engineering applications. Founder of Wisabi Analytics, focused on practical, production-ready AI content and tooling.