Case Studies

Each project taught me something I could not have learned any other way.

The hardest part of building AI for new infrastructure is that the data you need does not exist yet.

At 60,000 packages per day, 95% accuracy means 3,000 failures. The product question is always about trust, not precision.

In government, the product is not the dashboard. It is the decision the dashboard enables a Minister to make.

When your raw data is radio signals refracted through the atmosphere from low-earth orbit, the product challenge is making the invisible tangible.

Writing

Longer-form thinking on product, technology, and the spaces between.

Why the 500-year-old Neo-Confucian unity of knowledge and action is the best mental model for product management in hard environments.

The product management playbook for AI changes completely when your deployment environment has physics, safety constraints, and humans in the loop.

The best product decisions I have made came from embracing constraints rather than trying to engineer around them.