Forum Discussion

AlAmeenN's avatar
AlAmeenN
Frequent Visitor
3 months ago
Solved

Fabric Data Science standalone vs. hybrid with Azure ML for production MLOps

Hi, I'm working through an architectural decision for our team's production machine learning platform and would really value the perspective who have been through this in practice.   The core ques...
data science
  • v-aatheeque's avatar
    v-aatheeque
    3 months ago

    Hi AlAmeenN 
    Thanks for reaching out to Mircrosoft Fabric Community Forum.

    • If your use case is primarily focused on data preparation, model training, experiment tracking and batch scoring within the Fabric ecosystem, Fabric Data Science may be sufficient and has the advantage of keeping everything in a single platform.
    • However, if you're looking for more mature MLOps capabilities around model deployment, monitoring, governance, dedicated compute resources or advanced CI/CD workflows, it may be worth evaluating Azure Machine Learning alongside Fabric.
    • A key consideration is also how your workloads are expected to scale. Since Fabric workloads share capacity resources, organizations with larger or more demanding ML workloads sometimes whether dedicated Azure ML compute provides additional operational flexibility.

    For additional guidance, you may refer to the following documentation:
    Overview of Microsoft Machine Learning Products and Technologies - Azure Architecture Center | Microsoft Learn
    IDEAS journey to a modern data platform with Fabric - Microsoft Fabric | Microsoft Learn

     

    Hope this helps!!

    Thank You.

AlAmeenN's avatar
AlAmeenN
Frequent Visitor
3 months ago

Hi v-aatheeque 

 

Thank you for the message

 

Since most of our enterprise data already resides in Microsoft Fabric/OneLake, and at the same time we need to build a production-ready enterprise MLOps framework with capabilities such as CI/CD, deployment, monitoring, governance, and automated retraining, would a hybrid approach be the recommended architecture?

Specifically:

  • Microsoft Fabric for data engineering, feature engineering, and transformations

  • Azure Machine Learning for model lifecycle management, deployment, monitoring, governance, and CI/CD workflows

Also, are there any official reference architectures, implementation guides, or customer examples specifically for:

  • Fabric + Azure ML hybrid architecture

  • OneLake integration with Azure ML

  • Fabric pipelines/notebooks integrating with Azure ML deployment workflows

v-aatheeque's avatar
v-aatheeque
Icon for Community Support rankCommunity Support
3 months ago

Hi AlAmeenN 

Thank you for your follow-up.

For additional guidance around enterprise data platform architectures, MLOps capabilities, model deployment and lifecycle management, you may find the following official Microsoft documentation helpful : 

Analytics End-to-End with Microsoft Fabric - Azure Architecture Center | Microsoft Learn

MLOps machine learning model management - Azure Machine Learning | Microsoft Learn

 

Hope this helps!!

Thank You.

  • v-aatheeque's avatar
    v-aatheeque
    Icon for Community Support rankCommunity Support
    3 months ago

    Hi AlAmeenN 


    Have you had a chance to look through the responses shared earlier? If anything is still unclear, we’ll be happy to provide additional support.