For Heads-of · Practitioner
Model Deployment
The process of putting a trained AI model into production where it makes real-world predictions or decisions affecting users or business outcomes.
- deployment
- production
- infrastructure
- operations
Considerations
Choosing infrastructure (cloud, on-premise, edge), API design, latency and throughput requirements, security and access controls, monitoring and rollback capabilities.
Challenges
Models often perform differently in production than in testing due to data distribution shifts, novel user inputs, or adversarial attacks. Deployment requires robust monitoring and response protocols.
Governance Responsibility
Organizations are responsible for system behavior in production. They must verify the model performs acceptably, monitor for failures, and have procedures to quickly remove or patch failing models.