ML Ops Engineer
1 year contract
100% remote
Building the Azure AI Instance - This is the foundational infrastructure work that everything else depends on. The ML Ops Engineer is responsible for designing, provisioning, and maintaining the Azure environment that hosts all AI/ML workloads for Novolex.
Environment Setup: Provision Azure resource groups, networking (VNets, private endpoints), and identity management (Entra ID, RBAC). Configure Azure Machine Learning workspaces with appropriate compute targets (CPU/GPU clusters, serverless endpoints).
MLOps Pipeline Design: Build end-to-end ML pipelines using Azure ML Pipelines or Fabric Data Factory. Implement model training, evaluation, registration, and deployment workflows with full versioning and reproducibility.
Security & Compliance: Implement data encryption at rest and in transit, managed identities, key vault integration, and network isolation. Ensure alignment with Novolex IT security policies and the AI Governance Framework.
Cost Management: Monitor Azure consumption via Cost Management + Billing. Set budgets, alerts, and implement auto-scaling policies to optimize spend against the approved AI CoE budget.
Monitoring & Alerting: Configure Azure Monitor, Application Insights, and Log Analytics for infrastructure health, model drift detection, and pipeline failure alerting. Set up dashboards in Power BI or Azure Workbooks.
Data Integration & Analytics
Microsoft Fabric serves as the unified analytics platform connecting Novolex's data estate to AI workloads. The ML Ops Engineer bridges the gap between raw enterprise data and production-ready ML features.
Lakehouse Architecture: Design and build Fabric Lakehouses using medallion architecture (Bronze/Silver/Gold layers). Ingest data from SAP, Snowflake, Azure SQL, and flat file sources via Data Factory pipelines and Shortcuts.
Semantic Models: Create Power BI semantic models on top of Gold-layer tables to enable self-service analytics for business users while ensuring AI/ML pipelines consume the same curated datasets.
Notebooks & Spark: Develop PySpark and Python notebooks in Fabric for feature engineering, data exploration, and ad-hoc analysis. Leverage Fabric's built-in MLflow integration for experiment tracking.
Data Quality & Lineage: Implement data validation rules, freshness checks, and automated lineage tracking via Microsoft Purview integration. Flag and remediate data quality issues before they impact model performance.
Real-Time Capabilities: Evaluate and implement Fabric Real-Time Analytics (KQL databases, Eventstreams) for use cases requiring near-real-time data ingestion from manufacturing systems and IoT sensors.