We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
PyData Chicago: Building an MLOps platform at HelloFresh, by Dr. Erik Widman
Discover how HelloFresh built an end-to-end MLOps platform with Dr. Erik Widman's PyData Chicago talk. Topics include automation, tool selection, platform design, and the 10 core categories of MLOps solutions.
- MLOps is a systematic approach to deploying and maintaining machine learning models in production.
- HelloFresh created an end-to-end MLOps platform focusing on automation, portability, and scalability.
- Mapping the maturity of MLOps within the organization revealed consistent tooling, lack of standards, and scaling issues.
- The HelloFresh MLOps platform aims to address these pain points in a comprehensive manner.
- MLOps solutions address machine learning operations throughout the entire project life cycle.
- A scientific and structured tool selection process can simplify communication with C-level executives.
- An MLOps platform should offer flexibility and configurability, while providing simple and intuitive API classes.
- Consideration of user roles and feedback is essential to designing an effective MLOps platform.
- Automation of 80% of model creation processes is desirable, with the remaining 20% being simple to create.
- The 10 core categories of an effective MLOps platform are: data storage, model training, model validation, deployment, model monitoring, infrastructure, orchestration, endpoints and monitoring, integration layers, and simplification techniques.