Senior ML Engineer (Python / GCP)

Quick Facts

• Role: Senior ML Engineer (Python / GCP)
• Contract length: 12 months
• Work mode: Remote
• Start: 30 November 2026

Description

Build, deploy, and maintain production machine learning models that directly influence business decisions related to communication targeting, recommendations, and marketing budget optimization. Own the process end-to-end—from feature engineering and data quality checks in BigQuery, to deployment on GCP, and delivering predictions to external platforms via APIs for automated actions like campaign optimization, recommendations, or pricing. Tune and evaluate gradient boosting models on business-relevant metrics and calibrate predictions to understand how errors impact downstream decisions.

Responsibilities

• Build, deploy, and maintain production ML models in well-tested, reliable Python code
• Create features, run data quality checks, and perform scheduled data transformations in BigQuery
• Deploy and operate ML workloads on GCP (Vertex AI, Pub/Sub, Cloud Run, Composer/Airflow)
• Integrate predictions with external platforms via APIs to automate actions
• Tune and evaluate gradient boosting models using metrics important to the business
• Calibrate predictions and assess how model errors influence subsequent decision-making

Requirements

• 5+ years of practical ML engineering experience in Python, including responsibility for deployed models and reliable production code
• Experience integrating predictions with external platforms via APIs to enable automated actions (campaign optimization, recommendations, pricing)
• Advanced SQL and hands-on BigQuery experience: feature engineering, data quality controls, scheduled transformations
• Practical experience deploying ML workloads on GCP using Vertex AI, Pub/Sub, Cloud Run, and Composer/Airflow
• Production experience with gradient boosting models (LightGBM, CatBoost, or XGBoost), including tuning and evaluation against business metrics
• Ability to calibrate predictions and evaluate how prediction errors affect downstream decisions