Senior Machine Learning Operations Engineer
About this role
Building and operating real-time inference services that score models for the risk decision engine forms the core of the daily work, alongside establishing robust model deployment infrastructure and comprehensive observability systems. The Machine Learning Platform team at Mercury creates paved paths from trained models to reliable production deployments, enabling data scientists to register, deploy, and monitor models without carrying heavy operational burdens themselves.
As the scope of machine learning in risk decisioning rapidly expands to govern fraud and financial crime prevention, the need for scalable and reliable infrastructure grows increasingly critical. Engineers joining this group will partner closely with Risk Data Science colleagues to facilitate smooth development-to-production handoffs and implement advanced experimentation capabilities such as champion and challenger routing. Ideal candidates bring strong backend fundamentals, a collaborative mindset, and a genuine drive to shape a brand-new platform team supporting secure financial experiences for startups.
What you will do
- Build and operate the real-time inference service that scores models for the risk decision engine with low latency and high availability.
- Own model deployment infrastructure covering registry, versioning, CI/CD with performance checks, shadow mode, and staged rollouts.
- Construct model observability systems including availability, latency, error monitoring, and drift detection as a retraining trigger.
- Partner with Risk Data Science to manage models from initial development handoffs through active production operation.
- Implement experimentation capabilities such as champion/challenger and canary routing alongside explainability outputs.
What you need
- Bring five or more years of experience in machine learning engineering, backend software engineering, MLOps, or closely related fields.
- Demonstrate production ML service experience involving deploying, serving, and operating models in low-latency and high-availability contexts.
- Possess strong backend engineering fundamentals in Python utilizing API frameworks like FastAPI or Flask.
- Show practical experience with model deployment and lifecycle tooling including model registries, versioning, and staged rollout patterns.
- Display familiarity with the data layer supporting machine learning such as SQL, low-latency key-value stores, and streaming pipelines.
Nice to have
- Familiarity with a modern data stack including Snowflake, dbt, Dagster, or Airflow
- Experience operating in a regulated, audit-sensitive, or compliance-adjacent environment
- Exposure to functional languages or willingness to work across a stack that includes Haskell, React, and TypeScript
Benefits
- Base salary
- Equity including stock options and RSUs
- Comprehensive total rewards package updated regularly using reliable industry data
Skills and tech stack
Questions candidates ask
What locations are eligible for this position?
Candidates can work from San Francisco, New York, Portland, or remotely within Canada or the United States. New hire offers are made based on experience, expertise, and geographic location.
What is the compensation range for the role?
The target base salary range for US employees is $166,600 to $208,300 USD, while Canadian employees range from $157,400 to $196,800 CAD. The total rewards package also includes equity in the form of stock options or RSUs alongside benefits.
What technologies and tools does the team use?
The tech stack relies on Python with frameworks like FastAPI or Flask, alongside SQL, key-value stores like Redis or DynamoDB, and streaming pipelines such as Kafka, Kinesis, or Redpanda. Additional tools include modern data stack components and exposure to functional languages like Haskell.
What does the Machine Learning Platform team actually own?
The team owns the production machine learning lifecycle, encompassing systems that take a model from registry through deployment, real-time inference, observability, and retraining. They build the platform that lets data scientists register, deploy, and observe models in production.