MindWalk Holdings Corp. showed its ReefIQ platform publicly for the first time this year, the company said.1 Days later it filed patent EP26187897.9, adding a computational layer atop its foundational HYFT patent EP3881326A1.2
HYFT is built on roughly 660 million biological patterns — function-aware fingerprints that encode relationships between sequence, structure, and function.3 MindWalk's pitch: the representation of biology, not the model built on top of it, is the hard part and the place to build durable advantage.3
Biological data generated during drug discovery is scattered across files, formats, systems, and teams, MindWalk said. Before AI can reason about it, that data must be reconnected into a single, governed, queryable structure.1
The filing lands amid a broader buildout of AI-native infrastructure across pharma R&D. NVIDIA's BioNeMo, Basecamp's EDEN, Edison's Kosmos, Owkin's OwkinZero, and Boltz Lab are emerging alongside MindWalk's HYFT and ReefIQ as competing infrastructure layers. NVIDIA has anchored the buildout with hardware partnerships at Thermo Fisher and Eli Lilly.
Novo Nordisk is refocusing on its core GLP-1 franchise while an AI partner advances a Parkinson's program.4 The company is also pivoting away from owning cell therapy manufacturing in-house, licensing production instead to specialists like Cellular Intelligence. That shift coincides with accelerating stock gains and solid first-quarter earnings.4
The common thread: pharma majors are outsourcing infrastructure-heavy R&D functions — manufacturing, data representation, compute — to specialized AI vendors, while keeping core drug franchises in-house. MindWalk's patent filing bets that owning the biological representation layer, not any single foundation model, becomes the defensible position as AI platforms proliferate across drug discovery.
