Post-Training and Deploying Open Source Reasoning Models in Foundry | BRK232

Name: Post-Training and Deploying Open Source Reasoning Models in Foundry | BRK232
Uploaded: 2026-06-04T12:59:07.000Z
Channel: Microsoft Developer

Microsoft foundry Open Source reasoning models Post Training Reinforcement learning Model deployment Slime framework Verl framework Trl framework Rlhf Microsoft build 2026 Machine learning ops Llm fine Tuning

Microsoft Developer June 4, 2026

AI summary

This advanced session demonstrates how to use Microsoft Foundry to collect production traces, curate datasets, and post-train open-source reasoning models using reinforcement learning frameworks like slime, verl, and TRL. It covers when RL drives real gains in model performance and how to redeploy improved models without infrastructure management, using a retail customer service agent scenario for demonstration.