Founded in April 2025 in Abingdon by Pete Grindrod, Professor of Mathematics at the University of Oxford, Astut develops hybrid artificial intelligence technology designed to handle situations where conventional AI systems fail due to the absence of historical data. Combining symbolic reasoning and generative AI, Astut’s solution enables decisions that are explainable, traceable, and auditable in high-impact contexts such as defence, energy, and finance.
Based at the Culham Science Centre, the first UK government-designated “AI Growth Zone”, the company’s reasoning engine models possible choices using logical constraints and limited data. This allows it to address High Stakes Unseen Decisions (HSUD) — situations where organisations face unique, unprecedented choices with major consequences. The technology is being deployed through channel partners and embedded into their products and services for end-users operating in critical environments.
Astut raised €1.8 million in Seed funding, co-led by East X Ventures and Sure Valley Ventures (SVV), with participation from the UK Innovation & Science Seed Fund (UKI2S) managed by Future Planet Capital. The investment will accelerate product development and strengthen collaborations with key institutions including the UK Ministry of Defence, the UK Atomic Energy Authority, and partners across retail, consulting, and financial sectors.
The company operates within a rapidly expanding European DeepTech landscape focused on explainable and reasoning-based AI. During 2025, several players in this segment have secured funding, including Reshape Systems (€856k, Switzerland), Dragonfly (€3M, UK), EvoluteIQ (€44M, Sweden), and Optimuse (€4M, Austria). The UK continues to stand out as a key hub for sovereign and trustworthy AI innovation, attracting both private and institutional investment.
With this round, Astut aims to advance its hybrid AI platform toward a new generation of sovereign, explainable, and creative intelligence, capable of addressing complex decision-making challenges that current data-driven AI models cannot resolve.
