In the creation of a system that involves artificial intelligence (AI), the actual AI usually only takes a small fraction of the total amount of work, while a major part of the implementation entails the surrounding infrastructure, starting from data collection and verification, feature extraction, analysis, resource management, and serving and monitoring (David Sculley and others. Hidden technical debt in machine learning systems, 2015).
In this chapter, we'll deal with monitoring and model versioning, visualizations as dashboards, and securing a model against malicious hacking attacks that could leak user data.
In this chapter, we'll be covering the following recipes:
- Visualizing model results
- Serving a model for live decisioning
- Securing a model against attack
