How quantization-aware fine-tuning lowers cost barriers for domain adaptation and specialized assistants.