One of the many challenges faced in building AI solutions for healthcare is gaining clinical buy-in, convincing doctors to trust the ‘blackbox’. For them, it is a questionable science which they need to understand before trusting AI as part of their work. To bridge this gap, we have decided to expose the internals of our core algorithm in a scientific publication, in Nature Medicine. I will discuss why choose this path, what alternatives were considered? What’s painful about exposing your IP to the world? How to bridge the worlds of academia and industry research, build the story, prepare the code and experiments, and what workarounds we found to address seemingly impossible technical requirements. Finally, it’s retrospect impact, and how others can benefit from this story.