Bringing DevOps and Automation to Machine Learning via MLOps¶
The vast majority of organizations are new to AI/ML. As a result, most in-house systems and processes supporting this is likely ad-hoc. Industry analysts like Gartner forecast that organizations will need to quickly transition from Pilots to Production with AI/ML in order to make it across the chasm.
Most organizations already have reasonably mature DevOps processes and systems in place. So, going mainstream with AI should be a walk in the park. Correct? Turns out that this is not really true “IT leaders responsible for AI are discovering the AI pilot paradox, where launching pilots is deceptively easy but deploying them into production is notoriously challenging.” by Chirag Dekate, Gartner
In this blog, we will try and answer the following question:
Why do we need a new process called MLOps when most organizations already have reasonably mature DevOps practices? How is MLOps different from DevOps?
How is MLOps different from DevOps?¶
Both DevOps and MLOps are practices aimed at improving the efficiency and effectiveness of software development and deployment processes. Let's try to understand how they are similar and different.
DevOps (Development and Operations)¶
DevOps is a set of practices that combines software development (Dev) and IT operations (Ops) to shorten the system development life cycle while delivering features, fixes, and updates frequently in close alignment with business objectives.
MLOps (Machine Learning Operations)¶
MLOps is a practice for collaboration and communication between data scientists and operations professionals to help manage the production machine learning (ML) lifecycle. It aims to automate and streamline the continuous integration and delivery of ML models.
The image above shows the typical workflow that organizations have to implement for MLOps. As you can see, it looks very different from DevOps. Let us review the most important differences between MLOps and DevOps.
| Aspect | DevOps | MLOps |
|---|---|---|
| Nature of Artifacts | Deals with code artifacts like binaries, libraries, and executables. | Involves not just code but also data and models. Models change with data. |
| Versioning | Primarily focuses on versioning code. | Requires versioning of code, data, and models for reproducibility. |
| Testing | Emphasizes unit tests, integration tests, and system tests. | Includes data validation, model validation, and performance evaluation on datasets. |
| CI/CD | Automates building, testing, and deploying applications. | CI/CD pipelines are more complex, with data preprocessing, model training, validation, and deployment. |
| Monitoring and Maintenance | Focuses on application performance metrics like uptime. | Monitors model performance (accuracy, precision, recall) and tracks data and model drift. |
| Collaboration | Collaboration between developers and IT operations. | Involves collaboration between data scientists, ML engineers, and operations teams. |
| Infrastructure | Utilizes infrastructure for hosting apps and services. | Requires specialized infrastructure for model training (e.g., GPUs) and big data processing. |
Now since we understand the differences, let us look at this with the lens of a real life Rafay customer whose MLOps implementation allows them to rollout ~5 model updates/day.
In order to do this at scale, they have a few more user personas involved with their MLOps system. The image below provides a bird's eye view of Who does What? with a typical MLOps implementation.
Conclusion¶
In this blog, we discussed how DevOps is different from MLOps. Although there are shared concepts, it is critical for organizations to not conflate them as the same thing. It is important for organizations to address MLOps as its own first class citizen.
-
Free Org
Sign up for a free Org if you want to try this yourself with our Get Started guides.
-
Live Demo
Schedule time with us to watch a demo in action.
-
Rafay's AI/ML Products
Learn about Rafay's offerings in AI/ML Infrastructure and Tooling
-
Upcoming Events
Meet us in-person in the Rafay booth in one of the upcoming events