Together AI launches Together Post-Training, with its forward-deployed team helping customers run RL and SFT end to end

Together AI has launched Together Post-Training, a managed service that lets teams define their own reinforcement learning (RL) and supervised fine-tuning (SFT) training loops on dedicated GPU allocations. Together handles the underlying infrastructure. It supports LoRA and, for select models, full-weight training, and launches with open-weight models including GLM 5.3, Kimi K3 and Qwen 3.8-27B.

Forward deployed engineering is part of the offering itself. Together writes that many companies that could benefit from post-training "never start it, or start and stall", so its forward-deployed team works with customers on the end-to-end workflow. In the launch post, the founding engineer of recruiting-agent startup Jack & Jill says Together's team helped design its experiments and that its "forward-deployed engineers wrote the scripts that got us to first results".

Why it matters for FDEs: custom model training is becoming a hands-on customer engagement, not only a self-serve API. FDE work at AI infrastructure companies like Together increasingly means reward design, data and environment setup, evaluation and deployment alongside the customer. Together is hiring Forward Deployed Engineers focused on Inference & Post-Training.