Machine Learning Engineer II (Underwriting ML)

Affirm is hiring a Machine Learning Engineer II to join the Underwriting ML team, working remotely from Canada. You'll develop and deploy machine learning systems that power real-time transaction decisions, evaluating repayment risk and expected value across Affirm's checkout platform. Working alongside experienced ML engineers, data specialists, and cross-functional partners, you'll move models from concept through production, then maintain and improve them as user behavior and market conditions shift.

Responsibilities

• Develop and refine underwriting prediction models using techniques for tabular and sequential data
• Design, build, and scale feature pipelines and training datasets leveraging proprietary and third-party data sources
• Prototype novel modeling approaches and features; conduct offline experiments and validate performance before production rollout
• Integrate models into batch and real-time decision systems; optimize for reliability, latency, and operational robustness
• Establish monitoring and instrumentation for model and data health; define retraining and backtesting workflows
• Partner with Engineering, Risk Analytics, Product, and ML Platform teams to define requirements, evaluate tradeoffs, and communicate findings to technical and non-technical stakeholders

Requirements

• 2+ years of professional experience as a machine learning engineer, or a PhD in a relevant field
• Production-grade Python proficiency
• Experience building and evaluating classification models, particularly with gradient-boosted decision trees (LightGBM, XGBoost, CatBoost, or equivalent)
• Experience with a deep learning framework (PyTorch preferred)
• Experience with distributed data processing or parallel compute frameworks (Spark preferred; Ray, Dask, or similar)
• Experience with ML lifecycle tooling for training orchestration, experimentation, and model monitoring (Kubeflow, Airflow, MLflow, or equivalent internal platforms)
• Proficiency with AI-powered developer tools (Claude Code, Cursor, or similar) as part of regular development workflows