**Overview**

We’re looking for data scientists to help build the next generation of post-training methods for frontier models at Microsoft AI. You’ll join a small, high-impact team working across all stages of post-training, with a focus on evaluation design, high-quality training data, and scalable data pipelines for state-of-the-art foundation models.

In this role, you’ll help turn raw model capability into reliable, aligned, and measurable performance improvements, directly shaping how frontier models behave in real-world deployments.

About the Role:

Microsoft AI is building the next generation of frontier models that power Copilot and other large-scale AI experiences. The Post-Training team is responsible for transforming powerful pretrained models into robust, aligned, and high-performing systems used by millions of people worldwide.

Our work focuses on improving general quality, instruction following, coding and math ability, tool use,...