ML Ops Engineer

Coinflow•Chicago, IL

21h•$120,000 - $180,000

About The Position

Coinflow is hiring an MLOps Engineer to own the platform that moves fraud, risk, and payments models from development into production. The role builds pipelines, serving infrastructure, and monitoring to score transactions in real time and retrain models safely. Because models are part of the payment authorization path, latency, reliability, and explainability are critical alongside model accuracy. The engineer will use AI-native tooling to accelerate pipeline development, evaluation, and model operations. The position owns real-time model serving and safe fallbacks; end-to-end training, validation, and deployment pipelines; model versioning and controlled rollouts; production monitoring and drift detection; evaluation of third-party fraud and risk models; and data and scoring capabilities for merchant-facing risk tools. The role works with transaction, device, and chargeback data and supports merchants in identifying risky customers and configuring rules based on Coinflow's signals.

Requirements

At least 4 years of experience in MLOps, ML infrastructure, or ML platform engineering.
Experience running models in production with latency and uptime requirements.
Strong Python skills and sound production engineering practices.
Ability to read and write TypeScript/Node.
Hands-on experience with ML pipeline and serving tools such as MLflow, Kubeflow, SageMaker, Vertex, BentoML, Feast, or similar.
Infrastructure skills with containers, Kubernetes, CI/CD, and at least one major cloud provider.
Ability to build reliable feature pipelines using operational, warehouse, and streaming data, and understanding of training/serving skew.
Experience monitoring production models, including drift and calibration, and relating model metrics to business outcomes such as fraud loss and approval rates.
Familiarity with fraud, risk, or credit modeling and explaining model behavior to non-ML stakeholders such as banks and auditors.
Comfort using AI-native tooling such as Claude Code as a daily tool, or genuine excitement to begin using it.
A bias toward shipping practical solutions.

Nice To Haves

Experience in fintech, payments, or another regulated environment.
Experience with real-time inference in a transaction or decisioning path.

Responsibilities

Build and operate the real-time inference layer that scores transactions in the authorization path, meeting latency requirements and providing safe fallbacks so model issues do not disrupt payments.
Own training, validation, and deployment pipelines end to end, including reproducible feature pipelines with point-in-time correctness.
Implement model versioning, registries, shadow mode, champion/challenger testing, canary rollouts, evaluation gates, and one-click rollback.
Monitor model latency, score distributions, feature drift, approval rates, fraud, and chargeback outcomes; connect metrics to the observability stack and configure actionable alerts.
Integrate and evaluate external fraud and risk vendors using parallel and shadow testing, and compare their performance against Coinflow's data.
Provide data and scoring capabilities for merchant-facing risk products, enabling merchants to identify risky customers and set blocking rules using Coinflow signals.