Loading...
Mid LevelFull-timePythonGCPDockerMachine LearningPyTorchCI/CDML
Salary Range
$102,000 - $138,000 /year
This salary is estimated based on similar roles. The actual salary may vary.
About the Role
We are looking for an ML Engineer to join GoFasti, a Talent-as-a-Service company that connects developers and designers from Latin America with global companies. This role involves designing, building, and deploying end-to-end LTV prediction systems.
Key Responsibilities
- →Design, build, and deploy end-to-end LTV prediction systems, covering data ingestion, feature engineering, model training and evaluation, and deployment and monitoring
- →Develop ML approaches that work within adtech constraints such as delayed or sparse conversions, noisy attribution, and changing platform policies
- →Own MLOps, including reproducible training pipelines, CI/CD for models, logging, monitoring, and alerting
- →Perform data quality checks and drift detection
- →Collaborate closely with Product and GTM teams to translate business goals into model objectives
- →Help define and evolve the company’s Signal Engineering playbook
Requirements
- ✓Strong foundations in machine learning with the ability to reason from business objective, data limitations, model choice, and deployment
- ✓Hands-on experience building production ML systems, including training pipelines, deployment, and monitoring
- ✓Experience with LTV modeling, such as probabilistic or BTYD-style approaches, survival or retention modeling, regression/classification for value prediction, and model calibration
- ✓Comfortable working with modern data stacks and cloud environments
- ✓Core: Python, SQL, Docker (2–4 years experience)
- ✓Data & Warehouse: BigQuery, dbt-style transformations, event and transactional pipelines (Shopify, CRM, GA4, CDPs)
- ✓Cloud: GCP (Cloud Run, Pub/Sub / queues, scheduled jobs), secure APIs and services
- ✓Machine Learning: scikit-learn, XGBoost / LightGBM / CatBoost, optional PyTorch And MLflow or Weights & Biases for tracking
- ✓Orchestration: Airflow, Prefect, Dagster (approach matters more than the tool)
