Notebook-driven MLOps lab for domain-specific LLM fine-tuning: a 5-phase JupyterLab + boto3 walkthrough (setup, data prep, QLoRA train/merge/deploy, eval, testing) comparing a frontier model (Claude), a base open model (Qwen3), and a fine-tuned model on the same tasks with an LLM-as-judge, on AWS SageMaker + Bedrock. - View it on GitHub