PrizePicks
Machine Learning Platform Engineer
Atlanta, GA preferred, Remote (Remote)$155k–$185kmidAdded today
About this role
PrizePicks is seeking a Machine Learning Platform Engineer to design and operate scalable ML infrastructure that powers real-time decision-making across their sports betting platform. You'll bridge data science experimentation and production by building low-latency inference services, feature stores, and MLOps pipelines that handle millions of daily transactions.
What you'll do
- Design and build end-to-end ML infrastructure to transition experimental models into production-grade services
- Develop low-latency inference deployment systems serving model predictions in milliseconds for oddsmaking and risk analysis
- Create and optimize a centralized feature store supporting model training across multiple business domains
- Implement MLOps workflows including automated retraining, monitoring, and CI/CD with focus on data drift detection
- Collaborate with Infrastructure team on ML platform components and establish best practices for model deployment
- Own ML systems end-to-end in production including on-call support and incident response
What they're looking for
- Python (expert level)
- Streaming architectures (Kafka, Flink, Google PubSub)
- MLOps tools (SageMaker, VertexAI, Vector DBs, Graph Databases)
- Docker and Kubernetes containerization
- Low-latency system design (<100ms inference)
- Caching systems (Redis, Elasticsearch)
- Go, C++, or Rust (for high-performance layers)
- Real-time data processing and monitoring
Opens the official application on the employer’s site. No login required.
PrizePicks
PrizePicks operates a daily fantasy sports platform with a mobile-first experience. They are hiring Software Engineers (across levels II-III) skilled in TypeScript, React Native, and Ruby on Rails, as well as Data Engineers to build infrastructure supporting analytics and platform growth.
- Website
- prizepicks.com
Likely interview questions
- Describe a production ML system you've owned end-to-end—what were the main scaling challenges and how did you address them?
- How would you design a feature store to support models across different business domains while maintaining sub-100ms inference latency?