MLOps and the next level of machine learning model implementation
When we talk about analytics in our organisations, we usually go into the concepts of descriptive, predictive and prescriptive, as well as in the phase of technological and data maturation that make it possible.
All this, however, requires a great structural effort that does not necessarily imply the insertion of what is absolutely new, but also the adaptation of the processes that it already has, such as having a development area and software itself.
Here is where ML Ops becomes relevant. Data scientists and highly analytical teams find it extremely difficult to leave what they build in a productive environment: and that means operational and automated.
So, what is MLOps?
MLOps or ML Ops is a set of practices that aims to reliably and efficiently deploy and maintain machine learning models in production. The word is a compound of “machine learning” and the continuous development practice of DevOps in the field of software.
But, what differentiates it from DevOPS?
The process of building and deploying a model involves operations that are not included in the traditional software development process.
- Data (preparation, analysis and treatment)
- Training
- Model Evaluation
- Packaging
- Delivery (Deployment)
- Supervision (Monitoring)
And… What is the proposed process for MLOps?
It needs to cover the entire cycle, the end to end of operations
- Training
- Packaging
- Evaluation
- Deployment
- Monitoring
And the most used tools?
From the well-known TensorFlow Extended to Kubernetes we have:
What use examples are there?
In the most used cloud platforms:
Based on GCP –
Based on SageMaker –
Based on Azure –
What do you think about it? Did you know anything about MLOps?
—
If you want to contact us directly, send us a message.