MLOps Engineer

About this role

The AI team owns the models inside the product: resume understanding, candidate-to-role matching, and the assistive tooling recruiters use every day.

Own the path from a trained model to a served one: pipelines, registries, serving, monitoring and rollback. The goal is that shipping a model is as boring as shipping a service.

These models influence who a recruiter sees, which makes fairness a design constraint rather than a review step at the end. Every ranking change is evaluated for differential impact before it ships, and the team has standing authority to stop a launch. The input is genuinely hard — CVs in every format a human can invent, multilingual and code-mixed text, and documents photographed at an angle in bad light.

Primary responsibilities

• Build and run the machine learning platform — training infrastructure, experiment tracking, model registry and serving
• Make training reproducible, including data versioning, because an unreproducible result is not a result
• Own deployment: canary, shadow traffic, rollback, and the ability to compare two models on live traffic safely
• Build the monitoring for drift, data quality and performance degradation
• Manage GPU capacity and cost, which is a real constraint on what can be tried

Skills required

Day to day this role works with Python, Kubernetes, MLOps and Model Serving.

• Four or more years across ML and infrastructure
• Have operated model serving in production, including an incident
• Three or more years in ML infrastructure or platform engineering
• Training infrastructure, experiment tracking and model registry
• Reproducible training including data versioning