According to @_avichawla, a Claude-based studio now fine-tunes HF LLMs via AutoTrain, supports LoRA, quantization, and lets users chat with trained models.