Continuous delivery for Machine Learning, the future of MLOps
MLOps itself is a derivative of DevOps, the thought being that there is an entire industry that exists for “Ops” for normal software, and that such an industry will need to emerge for ML as well. But it hasn’t yet. Various technologies has made it easy for people to build predictive models, so people have lots of predictive models now. But to get value out of models you have to deploy, monitor, and maintain them. Very few people know how to do this, even fewer than know how to build a good model in the first place.
This talk will be dedicated to the plans of what is MLOps, what is cases and how it will develop and evolve into a new industry.
- Director of Engineering at DataRobot
- Started with building API for Machine Learning for 3 customers, now - millions of requests per day
- Writes on Python in free from meetings time