I co-founded Polaron, an Imperial College London spinout building AI infrastructure for materials engineering. Bringing new materials to market typically takes years of trial and error, slowing innovation, and increasing cost. We saw a real bottleneck in the characterisation and design of the microstructure of materials - often relying on human intuition, and slow manual processes. We use computer vision and machine learning to extract actionable insight from microscopy data, linking microstructure to performance. Our platform enables faster learning, and data-driven materials optimisation. We are backed by leading investors, have secured early commercial contracts, and are working with manufacturers to embed AI directly into real-world R&D and production workflows.