Member-only story
From Development to Deployment: Fundamental Practices in MLOps
The growing number of “Ops” terms might reduce the prominence of MLOps. Every engineer (Software Developer, QA Engineer, SRE, Machine learning engineer, Data Engineer, or Data scientist) should know the fundamentals of MLOps (machine learning operations). Simply put, it involves applying DevOps practices to machine learning projects or products.
Also, please check this, it describes the basic principles and purposes of Ops culture if you are new to DevOps, DataOps, and MLOps.
ML Engineering is 10% ML and 90% Engineering. In an overall ML project, the ML model code or algorithm constitutes 20%; the remaining 80% involves data ingestion, training configurations, training and model metadata maintenance, model versioning, model inference, and A/B testing.
ML Activities
The build includes model training, and the run includes model deployment and inference. However, there are surrounding activities that MLOps encompasses. These activities are divided as follows:
Build Part:
- Data Management: Handling data acquisition, cleaning, validating and storage. Note: Data is an important artefact of any ML Project
- Training Pipeline Management: Orchestrating the flow of data through the…