Product development means finding designs that meet several requirements at once: performance, cost, weight, safety and more. But the number of possible designs is enormous, and there’s no way to test them all. So engineers rely on experience to explore just one part of the design space, and if they can’t find a good answer there, they go back and search somewhere else. This kind of rework is common across the industry.

In fact, finding an answer that satisfies several conditions at once is something we’ve all been doing since childhood. Think back to choosing snacks for a school trip. The rules were strict: no more than 500 yen, nothing that melts like chocolate, and nothing that leaves messy wrappers behind. Within those rules, you weighed up price, quantity and the balance of sweet and savoury to pick the perfect bag of snacks. That is multi-objective optimisation in a nutshell.

In other words, Japanese people have been training as multi-objective optimisation experts since primary school.

So why is it so hard in real product development? Because unlike snacks, you can’t know how each design option will perform until you run a simulation or an experiment.

In this talk, we’ll introduce the approach developed by Secondmind, a UK-based engineering AI company, to solve this problem. Secondmind uses AI to find the region where all requirements are met first, an approach we call solution space exploration.

The AI starts by drawing a map of the entire development space from just a small number of simulations or experiments. It also shows which parts of the map are well supported by data and where risks still remain. This means engineers know where to look from the very start of development, eliminating the rework of searching again.

We’ll introduce two products in this talk.

The first is Secondmind for DSE, which supports concept design in the early stages of development. It visualises the range of designs that meet multiple requirements at once, and lets you instantly see the impact of changing any requirement. In the automotive industry, it has helped find a design that cut costs by 3,900 yen per vehicle without any loss in performance.

The second is Secondmind for Calibration, which supports the calibration of control systems. It finds optimal settings while dramatically reducing the number of tests needed. In the automotive industry, it has cut the number of experiments by up to 80%.

With both products, the AI doesn’t just give you an answer. It explains, with data, why a particular design or control value was chosen. That makes it easier to communicate the reasoning behind development decisions, not only in technical meetings but also in product planning and sales conversations, both internally and with customers.

This talk is for engineers, as well as anyone in product planning or sales who is interested in making development more efficient.