Develops and ships post-training methods for large language models (LLMs) used in Microsoft Copilot, focusing on data collection, evaluation, finetuning algorithms, and prototyping new capabilities. Collaborates with pretraining and product platform teams to deploy models and improve them based on user feedback.