SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 166,600 - 208,300 / annual
Mercury is a fintech platform providing banking, treasury, and financial workflow solutions for startups and technology companies. The Machine Learning Platform (MLP) team is building infrastructure to operationalize machine learning models in production, with a focus on fraud detection and financial crime risk decisioning.
In this role, you will own the production ML lifecycle—from model registry through deployment, real-time inference, observability, and retraining. You'll build and operate a low-latency, highly available real-time inference service that scores models for Mercury's risk decision engine. Key responsibilities include:
• Design and operate model deployment infrastructure with versioning, CI/CD pipelines that include performance and bias checks, shadow mode, and staged rollout patterns (canary, champion/challenger).
• Build comprehensive model observability systems: monitoring availability, latency, errors, and model-specific signals like data drift to trigger retraining.
• Implement experimentation capabilities and explainability outputs (e.g., SHAP attributions) for model decisions.
• Partner closely with the Risk Data Science team to ensure smooth handoff of models from development to production operation.
• Take strong product ownership of the platform, self-organizing on projects and helping shape a brand-new platform team.
You should have 5+ years of experience in machine learning engineering, backend software engineering, MLOps, or related fields. Production ML service experience is essential—you've deployed and operated models in low-latency, high-availability environments. Strong backend engineering fundamentals in Python (FastAPI, Flask) are required, along with hands-on experience in model deployment tooling, CI/CD for ML, and observability/alerting for production services.
You're comfortable with the data layer: SQL, low-latency stores (Redis, DynamoDB), and streaming platforms (Kafka, Kinesis, Redpanda). Nice-to-haves include familiarity with modern data stacks (Snowflake, dbt, Dagster, Airflow), experience in regulated/compliance-sensitive environments, and exposure to functional languages or a polyglot stack (Haskell, React, TypeScript).
About Mercury
Fintech — banking, treasury, and financial workflow platform for startups and technology companies.
