Why do a significant chunk of data science projects never make it out of the lab, when AI’s business value comes from deploying machine learning (ML) models operationally? One big reason is the hype-driven motivation behind many AI initiatives, which often overlooks concrete use cases, business objectives and measurable outcomes. Another reason, though, is the lack of tooling and well-adopted practices around ML operations (MLOps).
ML development is a form of software development, and it ought to be pursued with the same rigor. Automation and practices around model training, testing, packaging, release, and even rollback are key. And once the model is deployed, monitoring it, retraining it, and testing for data drift are important too. An ad hoc machine learning workflow won’t provide this, but a full ML platform with true MLOps capabilities does.
If this sounds appealing, but you’re not sure how or where to start, join us for this free 1-hour webinar from GigaOm Research. The webinar features GigaOm analyst Andrew Brust and special guest Katie Gross from Dataiku, a leader across the entire AI lifecycle.