Custom Model Fine-Tuning

Fine-tune Llama, Mistral, Qwen or vendor models on your domain data. SFT, DPO, RLHF as appropriate. Eval harness — accuracy, latency, regression suite — included.