“Developing and deploying ML systems is relatively fast and cheap, but maintaining them over time is difficult and expensive” – D. Sculley et al., “Hidden Technical Debt in Machine Learning Systems, NIPS 2015
With an adoption that doubled compared to 2017 based on the McKinsey State of AI report, AI is going through a shift. Initiatives are going past the experimentation phase and making it to production. Some noteworthy changes in the space include:
- An increasing number of capabilities being used by organisations as part of their AI initiatives.
- An increasing level of investment is allocated to machine learning projects, which goes hand in hand with a higher adoption rate.
- More interest in collection, governance and ethics, aiming to ensure compliance for production deployments.
How do you navigate these changes in the fast-paced world of AI? The answer lies in building a well-oiled MLOPs practice. Much like DevOps changed the world of software development, MLOps will enable maturity for AI.
This guide offers some guidance for data scientists and ML practitioners who are looking to take their AI models from the experimentation phase to the production stage with MLOPs. Topics covered in this extensive guide include:
- What is MLOps?
- Benefits of MLOps
- MLOps lifecycle
- MLOps tooling
- Open source MLOps
The guide is based on the machine learning lifecycle. You will find MLOps principles and best practices to ease the life of data scientists and machine learning engineers in your team.
If you are looking to scale your AI initiatives and adopt machine learning operations (MLOps) as a practice, this guide will help you move forward.