Enterprises and startups alike are moving past the novelty phase of artificial intelligence (AI) and machine learning (ML) into a mature phase of implementing it in a production capacity for real business. If you’re going to use AI/ML to compete and win, you’ll need to optimize development productivity and make deployment, monitoring, and iterative improvement as frictionless and automated as possible. It’s no longer about getting ML to do cool stuff; it’s about making it do so repeatedly, reliably, and efficiently. In other words, it’s about production ML.
How can production ML be meaningfully and reliably implemented? Ironically, it’s as much about what comes before and after the training and testing of models as it is about the hardcore ML work itself. Data quality, data pipelines, ground truth monitoring, and model/prediction explainability are critical to ML. Getting there requires a full platform that supports data engineers and business users, in addition to data scientists and ML engineers. And beyond platform and tooling, production ML requires adopting the right philosophy and approach to the work and its business application.
To learn more about the philosophy, technology, and methodology of production ML and some of the pitfalls to look out for, join us for this free webinar from GigaOm Research. It features GigaOm analyst Andrew Brust and special guest Santiago Giraldo from Cloudera, the enterprise data platform company.