I work with CTOs who are frustrated with AI systems that work in the lab but fail in production.

Many AI initiatives fail because they rely on assumptions that do not hold up under real usage. For example, assumptions on the type of queries or inputs that the system will need to handle. Production success takes hard work, combining creativity, critical thinking, iteration and relentless testing.

My perspective was shaped by building systems where incorrect outputs carry real consequences. After my PhD at the Technical University of Denmark, I co-founded a company building AI for vessel optimization in live maritime operations. Our models were deployed into production environments where errors were immediately visible, operationally costly. I stayed through acquisition, serving as CEO, CTO, VP of AI.

That experience taught me that while building robust models is difficult, building robust AI systems is harder. And that modern AI solutions are complex systems that won't hold up under pressure simply because they use the latest LLM.

Since then I have:

• Built property valuation AI systems used by banks to meet Basel III requirements
• Architected AI solutions for SaaS companies operating in highly regulated markets
• Designed systems where traceability, evaluation discipline, and decision reconstruction were first-class architectural concerns

My role is typically:

• Leading architecture workshops
• Designing production roadmaps
• Fractional Head of AI leadership

What I bring:

• Experience shipping AI into live operations
• Deep understanding of how risk maps to technical design
• Creative systems thinking
• Direct, accountable collaboration

I work with only a few clients at any one time. It is important that we begin by making sure that we are a good fit for collaboration, because that is the best way of ensuring the best outcomes.