Magic systems like machine learning, neural networks, genetic algorithms, statistical analysis and content recognition are difficult to run in production. These systems are not deterministic and heavily dependent on real data. This makes it a challenge to test and deploy these systems with confidence.

This talk is about the requirements and demands of magical systems. How do you make miracles show their work.

This talk is based on my blog https://www.yasendinkov.com/posts/magicinproduction/ and a series of internal showcases. It's not tied to any particular technology. It's targeted towards operators and engineers that create AI solutions for end-users.