From the course: MLOps Essentials: Model Development and Integration
Selecting ML projects
From the course: MLOps Essentials: Model Development and Integration
Selecting ML projects
- [Instructor] In this chapter, we will explore the requirements and design aspects of machine learning, and how it helps create effective ML Ops processes. First, we start with how we should select ML projects for execution. The eventual goal for any machine learning project is to improve business outcomes. This means that the ML project should either help in increasing sales or reducing costs. An organization would and should only invest in an ML project if there are business benefits that outweigh the cost of building the model. During the exploration and experimentation stages, a number of alternate use cases are identified and analyzed. How does an organization choose projects to go forward with execution? What are the criteria for selecting ML projects? Why should they build one, and not just by an equal in service from the market? We should first start with whether the model could bring core business value to the organization. ML projects are expensive, and unless they create strategic differentiation to the product or service, they may not have long-term value. The next criteria is the availability of training data, including labels as required. Without good training data ML, projects will fail. A technology ecosystem should also exist in the specific domain for machine learning. This includes algorithms, libraries, frameworks, and pre-trained models as needed. For example, if your use case is in computer vision, then related-based technologies should be available and affordable. Next, comes the budget that is available to create a team. If ML projects are understaffed, they will take very long cycle times, and may outrun their usefulness. Time to market is critical, as the organization doesn't want to wait for years before the ML model is operational. Competition may overrun and beat them. Finally, ML projects do carry a risk of failure. If the model cannot meet desired performance requirements or cost requirements. So there should be an appetite for failure in the organization. Choosing the right project at the beginning is critical to maximize the probability of success at the end.
Practice while you learn with exercise files
Download the files the instructor uses to teach the course. Follow along and learn by watching, listening and practicing.
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.
