CuspAI's Namira platform designs materials that don't exist

Max Welling, machine learning professor at the University of Amsterdam and CTO of CuspAI, describes his company's platform as a search engine for materials that do not yet exist. On the Dutch podcast AIToday Live, he walks through how the system, called Namira, generates millions of candidate materials with AI, filters them for stability and practical use, and only then hands survivors to a lab for validation.

"Namira: een zoekmachine voor niet-bestaande materialen die met AI miljoenen kandidaten genereert, filtert op stabiliteit en valideert in het lab," reads the show description.

The same description credits the approach with pushing perovskite solar cells "richting de 30% efficiëntie" — toward the 30% mark — with lab measurements fed back into the models. Welling also argues that regulatory and organisational processes are a bigger drag on materials AI than the modelling itself, and that verification, not step-by-step interpretability, is what earns trust in the generated candidates. The episode lists AMOLF and Nobel laureate Geoffrey Hinton among the reference points; it does not spell out any specific industrial partner running the validation loop.