Senior Data Platform Engineer(Microsoft Fabric,MLOps)

Requirements

• These are the hard requirements. They are engineering requirements, not machine learning requirements.
• 5+ years building and operating production data platforms on Microsoft Azure.
• Hands-on Microsoft Fabric experience: workspaces, OneLake, Lakehouse and Warehouse structures, Fabric Notebooks, Fabric RBAC, and deployment pipelines. Strong Synapse, Databricks, or Azure Data Factory backgrounds are welcome where the candidate has already moved onto Fabric or can credibly show they will.
• Strong Python and SQL, with production-grade code structured for reuse, testing, and scheduled execution rather than exploratory notebooks alone.
• CI/CD applied to data and notebook assets using Azure DevOps or GitHub Actions, including automated unit and integration testing.
• Orchestration and scheduling experience with dependency management and pipeline-level failure visibility, using Fabric Data Pipelines, Azure Data Factory, Airflow, or equivalent.
• Working knowledge of a data quality framework such as Great Expectations, Soda, or an equivalent rules-based validation approach.
• Solid grasp of Azure identity and security: Entra ID, service principals, managed identities, Azure Key Vault, and RBAC. Specifically, experience making scheduled workloads run headless with no user-bound authentication.
• Understanding of Power BI consumption patterns against lakehouse and warehouse tables, sufficient to validate and document downstream access.
• Clear written English and the ability to produce runbooks and design documentation that a client operations team can follow without the author present.
• Minimum four hours of daily overlap with US Eastern business hours for standups, design reviews, and milestone demonstrations.
• Bachelor's degree in Computer Science, Data Science, Engineering, or a related field.
• Machine Learning Operationalization
• Working-level capability is required here. Deep specialization is not, and a modeling background is not.
• Experience taking at least one machine learning model into scheduled production and keeping it running. The model itself may have been built by someone else.
• Working understanding of model versioning and reproducibility, meaning what it takes to tie a production run back to a specific combination of code, configuration, environment, and data.