Machine Learning Operations (MLOps) Engineer

Position Summary

The MLOps Engineer builds and operates the platform and pipelines that take machine learning and AI models from development into reliable production use. This role owns model deployment, automation, monitoring, and lifecycle management, and serves as the ongoing owner of the enterprise sales forecasting models. Beyond operating the platform, the MLOps Engineer is expected to work inside the models themselves, diagnosing accuracy degradation and applying data science judgment to retune features, algorithms, and model logic so that forecast performance continues to deliver measurable business value. The MLOps Engineer partners closely with Data Science, Data Engineering, and Security teams to bridge experimentation and production, and is accountable for the operational health of models once they are live.

Key Responsibilities

• Design, build, and maintain CI/CD pipelines for model training, validation, deployment, and rollback.
• Productionize models developed by data scientists and vendor partners, converting experimental code into scalable, maintainable, and reproducible pipelines.
• Own ongoing maintenance and performance of the enterprise sales forecasting models, including accuracy tracking, periodic recalibration, and continuous improvement.
• Diagnose forecast accuracy degradation and apply data science techniques to improve results, including feature engineering, algorithm selection, hyperparameter tuning, and adjustment of model logic.
• Evaluate forecast performance using appropriate accuracy measures such as WMAPE, and report results and improvement trends to business stakeholders.
• Translate business context, including promotions, seasonality, café openings and closures, and operational disruptions, into model features and adjustments so forecasts remain useful for planning.
• Partner with Finance, Operations, and Supply Chain to understand how forecasts are consumed and ensure model changes deliver measurable business benefit.
• Deploy and manage models on Vertex AI, including training jobs, endpoints, batch prediction, and pipeline orchestration.
• Operationalize LLM and generative AI workloads, including in-warehouse model invocation from BigQuery, prompt and model version management, and output validation.
• Implement model monitoring for accuracy, drift, data quality, latency, and failure, with automated alerting rather than manual review.
• Build automated retraining workflows with clear promotion criteria and approval gates before a new model version reaches production.
• Establish model registry, versioning, and lineage practices so that every production prediction can be traced to a model version, code commit, and training dataset.
• Manage feature pipelines and ensure consistency between training and serving data to prevent training-serving skew.
• Monitor and optimize compute and inference cost, including model selection, batch versus real-time tradeoffs, and capacity utilization.
• Define and enforce infrastructure-as-code standards for ML environments, including environment promotion from development through production.
• Partner with Security and Privacy teams to ensure models and training data meet data classification, access control, and retention requirements.
• Support model governance and audit requirements by documenting model behavior, validation results, and approval history.
• Troubleshoot production model and pipeline failures, drive root cause analysis, and implement preventive controls.
• Collaborate with Data Engineering to define upstream data requirements and ensure pipeline reliability for model inputs.
• Contribute to architecture reviews and present ML platform designs and operational readiness for approval.
• Mentor engineers and data scientists on production standards, deployment practices, and operational discipline.