ML Ops Engineer

Please note:

This role is contingent upon a contract award. While it is not an immediate opening, we are actively conducting interviews and extending offers in anticipation of the award.

The Work At ICF our Digital Modernization Division is an information technology and management consulting organization that delivers integrated, strategic solutions to federal clients. We bring expertise in cloud, cybersecurity, enterprise architecture, data modernization, and digital transformation to support mission-critical government programs.
Join a team accelerating the modernization of a large federal agency's enterprise data and analytics ecosystem. This cloud-based platform provides data storage, analytics, governance, and AI/ML capabilities that enable thousands of users to transform data into actionable insights. As demand continues to grow, the team is focused on migrating legacy workloads, streamlining onboarding and support, expanding platform capabilities, and helping organizations across the enterprise adopt modern data and AI solutions  ML Ops Engineer will develop and maintain machine learning pipelines, model registries, deployment frameworks, monitoring solutions, and automated workflows that support enterprise AI initiatives.

Working closely with data scientists, data engineers, cloud architects, platform engineers, and Trustworthy AI stakeholders, this role enables the successful operationalization of machine learning capabilities across cloud-based environments.

Job Location:

Remote, however, strong preference for candidates who live in the Washington DC Metro Area. There will be occasional onsite meetings on the client site in Washington, DC.
* If you accept this position, you should note that ICF does monitor employee work locations, blocks access from foreign locations/foreign IP addresses, and prohibits personal VPN connections.

What You'll Do Design, build, and maintain production-grade MLOps pipelines for model training, evaluation, validation, registration, deployment, monitoring, retraining, rollback, and retirement. Develop reusable MLOps and LLMOps patterns using Azure Databricks, Databricks ML, MLflow, Unity Catalog, Azure Machine Learning, Python, and approved Azure AI services. Implement experiment tracking, model versioning, model registries, lineage, quality metrics, model cards, and approval workflows using MLflow and Unity Catalog to ensure models are discoverable, reusable, and auditable.

Build automated CI/CD pipelines using Git Hub Actions, Azure Dev Ops, Databricks Asset Bundles, and Terraform to promote models, code, configurations, and supporting infrastructure consistently across development, test, and production environments. Deploy and maintain scalable model-serving infrastructure for batch, streaming, real-time, and API-based inference workloads using Databricks Model Serving, Azure Machine Learning, Databricks compute, serverless capabilities, and GPU-enabled configurations, as applicable.

Establish automated testing and evaluation gates for model performance, data quality, fairness, safety, explainability, robustness, privacy, and release readiness. Implement model and feature monitoring, drift and bias detection, logging, tracing, alerting, and operational dashboards using MLflow Model Monitoring, Lakehouse Monitoring, and the broader platform monitoring stack. Develop automated retraining, champion/challenger promotion, rollback, and model-retirement workflows that prevent performance degradation and support reliable production operations.

Integrate ML workflows with source control, infrastructure provisioning, secrets management, identity and access controls, environment management, and enterprise release processes. Implement Trustworthy AI controls, including model documentation, impact assessments, audit logging, content safeguards, human oversight, and evaluation evidence aligned with Government-approved standards and the NIST AI Risk Management Framework. Partner with data scientists, AI engineers, data engineers, platform engineers, Dev Sec Ops , Security, and Governance personnel to make models repeatable, testable, deployable, and operationally supportable.

Monitor compute and serving utilization, tune cluster and endpoint configurations, apply tagging conventions, and implement cost-optimization strategies for training and inference workloads. Troubleshoot failed pipelines, deployment issues, model-serving incidents, performance degradation, and monitoring alerts while…