AIceberg – Enterprise AI Risk Management

Designing the control layer for AI agent testing, security, and compliance.

Project overview

AIceberg builds a platform that lets teams test, monitor, and validate the quality and security of their AI agents, whether built in-house or brought in from elsewhere. As AI agents moved from experiments into real products, the need for that kind of ongoing oversight grew with them, and AIceberg set out to give teams a clear, structured way to keep watch over systems that are often difficult to fully predict.

Role and collaboration

I worked alongside a product owner who also carried the product strategy for the platform, and I covered every page, feature, and flow in the product myself. We worked in a tight loop together: ideas got shaped into prototypes, those prototypes were tested with a small group of clients and mostly validated internally, and what came back fed straight into the next round of design.

What I built

Over close to two years, I built and refined a series of features this way, taking the dashboard through several redesigns as real usage and feedback made clear what needed to change. The result was a product that kept improving, not a single fixed design delivered once and left alone.