ℵ₁₁

Many organisations are currently developing AI proofs of value (PoVs) and minimum viable products (MVPs). However, these initial models too often become permanent, default solutions without a proper assessment of costs, business value, or user impact. While early enthusiasm is enough to kick off an AI project, transitioning beyond PoVs and MVPs requires a clear roadmap, the right talent, and an experimentation mindset. In one of my roles as a solution architect, I created a roadmap to incrementally roll out AI systems, evaluated the user and business impact of each roll-out, and adjusted the strategy based on new empirical findings from dozens of online and offline experiments.

A successful AI strategy requires an experimentation platform that helps your organization move from PoCs and MVPs to more valuable solutions and stay a step ahead of your competitors. Over the past few years, I have evaluated hundreds of AI models and helped build experimentation platforms that have allowed us to accelerate model assessment offline and through A/B tests. I have experienced the rise of LLM-as-a-judge and learned how it can speed up and deepen the insights gained from controlled and natural experiments.

In previous roles, I designed robust data pipelines—spanning from initial data collection to database integration, and delivered actionable insights to both business and engineering stakeholders. Additionally, I served as a Scrum Master and presented my work at various European conferences and industry meetups.