Valencia-based Maisa, a startup creating a novel agentic AI system that provides a ‘Chain-of-Work’ (CoW) to hint and confirm its outputs, has raised €4.7 million in a pre-seed funding spherical.
Led by NFX, the spherical noticed participation from Village International—backed by tech giants Mark Zuckerberg, Eric Schmidt, and Jeff Bezos—in addition to Sequoia scout and DeepMind PM Lukas Haas.
David Villalón, CEO and Co-founder of Maisa, emphasised the importance of their innovation: “Our launch of the Vinci KPU is critical within the growth of agentic AI for companies and for builders of AI functions and automation options. That is the primary time agentic AI turns into clear and auditable, as an alternative of the present ‘black field’ that obscures its outcomes and represents an enormous danger for companies. We’re delighted to have gained the backing of such legends in international tech growth and leaders of a few of the largest companies on the earth, not least for the popularity that their funding offers to the importance of what we’re constructing.”
Based in 2024 by David Villalón and Manuel Romero, Maisa has developed the Vinci Information Processing Unit (KPU), “world’s first agentic AI system“. The Vinci KPU makes use of a novel CoW strategy to make sure traceable, auditable, and deterministic outputs.
Not like conventional AI programs, which depend on probabilistic ‘next-word’ prediction and sometimes operate as opaque ‘black packing containers’, the Vinci KPU executes reasoning step-by-step utilizing attributable knowledge sources. This innovation eliminates hallucinations and introduces transparency, making AI viable for complicated business-critical functions.
Maisa’s system is designed to autonomously pursue targets, make choices, and adapt to dynamic environments whereas retaining people in management.
Early adopters of the Vinci KPU embrace a worldwide automobile producer enhancing provide chain resilience, a global oil and gasoline agency managing evolving compliance necessities, and a European client credit score establishment optimising its mortgage approval processes beneath strict regulatory scrutiny.
Based on Maisa, the know-how has confirmed itself towards trade benchmarks, rating alongside OpenAI’s o1 mannequin in graduate-level reasoning (GPQA) whereas outperforming main massive language fashions in coding (HumanEval), procedural adherence (ProcBench), and mathematical drawback fixing.
Maisa’s capacity to offer full auditability is shaping it to be an economical and scalable resolution for automation.
Anna Piñol, investor at NFX, commented on Maisa’s transformative potential: “Maisa goes to be a serious participant in RPA 2.0 serving to companies internationally remodel their core, business-critical capabilities by way of AI. It should enable them to work sooner, extra effectively and obtain new and radical methods of working. This launch will probably be seen as a pivotal second in AI growth this 12 months and as AI’s coming of age in sensible, business-enhancing use.”
Max Kilberg, investor at Village International, added: “David and the Maisa workforce are constructing a transformative know-how to show AI brokers into precise staff which can be able to reasoning by way of complicated workflows. We’re tremendous thrilled to be part of their journey and are very excited to see the brand new benchmarks and enterprise traction.”
With workplaces in Valencia and San Francisco, Maisa is positioning itself as a rising title in agentic AI. The funding will probably be used to reinforce the Vinci KPU’s capabilities and strengthen its go-to-market technique, concentrating on sectors like monetary providers, provide chain, and analysis.