Machine Learning Engineer

Lumenalta partners with forward-thinking organizations to build technology solutions that scale and accelerate business growth. They are seeking an experienced MLOps Engineer responsible for operationalizing machine learning at scale on the Databricks platform, bridging data engineering and ML to build infrastructure and workflows for reliable production deployments.

Responsibilities:

• Design and maintain MLflow-based workflows for experiment tracking, model registry, versioning, and lifecycle management
• Build and manage Feature Store infrastructure to enable reusable, consistent feature pipelines across teams and use cases
• Develop model deployment pipelines, including serving infrastructure, A/B testing support, versioning, and rollback strategies
• Implement CI/CD pipelines tailored for ML workflows, including automated testing, validation gates, and deployment triggers
• Orchestrate distributed model training on Databricks, optimizing for compute efficiency, reproducibility, and cost
• Monitor deployed models for data drift, performance degradation, and system health, triggering automated retraining workflows as needed
• Collaborate with Data Scientists and Data Engineers to reduce friction between experimentation environments and production

Requirements:

• 3–5+ years in MLOps, ML platform engineering, or DevOps for ML, with proven production ML deployments
• Hands-on expertise with MLflow for tracking, registry, and project management within Databricks or standalone environments
• Experience building and consuming Feature Store solutions (Databricks Feature Store or equivalent)
• Proven experience deploying and serving ML models at scale, including real-time and batch inference patterns
• Ability to design automated pipelines for model training, validation, and deployment using modern CI/CD tooling
• Strong familiarity with Databricks for distributed training, job orchestration, and cluster management
• Knowledge of model monitoring practices, including drift detection, alerting, and retraining triggers