Machine Learning Engineer - Inference / Serving

The role

You will design, optimize, and operate systems that bring Behavioral AI models to life in real time. You will work on turning trained models into performant, reliable, and continuously improving services for open-web and CTV products.

What you will do

• Design, optimize, and operate systems that bring Behavioral AI models to life in real time
• Turn trained models into performant, reliable, and continuously improving services
• Shape how models are packaged, versioned, rolled out, and observed across environments
• Ensure predictions are fast, accurate, and accountable
• Build or scale production ML serving systems
• Handle versioning, rollouts, rollback strategies, and live experimentation
• Optimize inference through model graph optimization, quantization, caching, batching, and efficient feature retrieval
• Write robust, high-performance code
• Monitor drift, track model lineage, and ensure observability
• Work with researchers to make models more deployable

Required qualifications

• Deep expertise in model deployment
• Experience building or scaling production ML serving systems
• Knowledge of versioning, rollouts, rollback strategies, and live experimentation
• Understanding of model graph optimization, quantization, caching, batching, and efficient feature retrieval
• Ability to write robust, high-performance code in Go, Rust, C++, or Java
• Comfort bridging to Python for model integration and analysis
• Ability to monitor drift, track model lineage, and ensure observability
• Knowledge of making serving systems reproducible and portable
• Applied ML understanding to reason about model performance and interpret trade-offs

Benefits

• Competitive Base Salary
• Meaningful equity & financial upside
• Annual bonus target based on personal and company performance
• Health, Dental, Vision available
• Unlimited PTO
• 401k with company match %
• Generous parental leave policy

About the company

Yobi is a Behavioral AI company that builds foundation models of human behavior grounded in real-world actions. They focus on bringing the performance of closed-web user acquisition to the open web and connected TV.