Building Sovereign AI: Yang Li, CEO of Cosine on AI Models, Coding Agents and the UK's Frontier AI Future

What does it take to build a frontier AI model lab in a world dominated by OpenAI, Anthropic and Google?In this episode of Riding Unicorns, James and Hector sit down with Yang Li, Co-Founder & CEO of Cosine, one of the UK's leading frontier AI companies building sovereign AI models for enterprise and government.Cosine began by fine-tuning foundation models for software engineering before evolving into a full frontier AI lab, training its own models from scratch. Today, the company works with organisations across defence, financial services and critical national infrastructure, deploying AI securely inside highly regulated environments.The conversation explores why sovereign AI has become a strategic priority, how smaller model labs can compete with the biggest players, and what the future of enterprise AI will look like.Yang also shares how AI is changing software engineering, why synthetic data has become a competitive advantage, and the lessons he's learned building one of Europe's most ambitious AI companies.Topics Covered• How Cosine evolved from AI coding tools into a frontier AI model lab • Why sovereign AI is becoming critical for governments and enterprises • Competing with OpenAI and Anthropic without billions in funding • Building AI models using synthetic data and post-training techniques • Why defence, banking and healthcare are driving enterprise AI adoption • The future of software engineering in the age of AI coding agents • Air-gapped AI, on-premise deployment and enterprise security • The geopolitical race for AI infrastructure and compute sovereignty • Hiring world-class AI talent and building high-performance teams • Founder lessons on conviction, resilience and staying alive long enough to winThis is a conversation about the future of frontier AI, the growing importance of sovereign technology, and what it takes to build an AI company that competes on the global stage.