Module 3: Understanding the AI Development Lifecycle: Implement Flashcards
What is the primary activity of the Implementation Phase?
Deploying an AI model into production.
What are the key steps in the AI system implementation phase?
- Perform readiness assessments.
- Deploy the model into production.
- Monitor and validate the model.
- Maintain the model.
What is the primary activity of the Implementation Phase?
Deploying an AI model into production.
What steps should an organization take during the implementation phase?
- Perform a readiness assessment.
- Set up continuous monitoring.
- Define a baseline to measure future iterations.
- Maintain the model to avoid model drift.
What does deploying an AI model involve?
Transitioning from a development and testing environment to a real-world, operational setting to be used for its intended purpose.
What are key considerations for AI model deployment?
1) Choose a deployment environment.
2) Package the model.
3) Make the model accessible.
Name 3 of the most common deployment environments and their advantages and disadvantages.
1) Cloud-based: A third-party cloud provider hosts the model and handles infrastructure.
- Advantages: easy to scale up or down; reduces the need to invest in hardware
- Disadvantage: potential latency issues and security risks could be introduced when a third party is handling the data
2) On-premise: The model is hosted on servers and hardware that is owned and managed by your organization.
- Advantage: greater control over deployment infrastructure, which is especially important if you handle sensitive data or are in a regulated sector
- Disadvantage: may require a greater up-front investment in hardware as compared to cloud-based environments
3) Edge: The model is hosted on “edge” devices like smartphones.
- Advantage: may have decreased latency and greater data privacy
- Disadvantage: the model may be limited by the edge device’s hardware, which can then limit the model’s computational power
What is packaging of an AI model and what is a common option for packaging?
Packaging creates a format for an AI model to be deployed.
A common option is “containerization,” which involves packaging the model and dependencies (i.e., everything the model needs to run effectively) into a self-contained unit. Containers can help reduce compatibility issues and make it easier to deploy the model in different environments (e.g., development or testing).
How is an AI model made accessible for real world use?
By allowing systems or applications to interact with it. (AKA “exposing the model”)
