Difficulty Onboarding New Developers Solved: ChatGPT Prompts for Consulting ML Engineers (Beginner)

You are a senior ML engineer with 10 years of experience in consulting environments, machine learning model deployment, and developer onboarding for client-facing data science teams. Help me write a deployment script so I can reduce production bugs. My situation: ML feature being deployed: [e.g., recommendation model update / sentiment analysis API / batch prediction pipeline for client reporting] Consulting client environment: [e.g., AWS SageMaker / Azure ML / on-premise server with restricted access] Onboarding problem: [e.g., new developers do not know which environment variables to set / the deployment process is undocumented and only one person knows it / each developer follows a different deployment order and introduces different bugs] Current deployment method: [e.g., manual SSH commands / partial shell script with no error handling / Jupyter notebook run in sequence] Production bug pattern: [e.g., model deployed without the correct dependency version / environment variable missing in production causes silent prediction errors / deployment succeeds but monitoring is not activated] New developer technical level: [e.g., junior ML engineer with Python experience but no DevOps background / data scientist with no deployment experience] Deployment frequency: [e.g., once per sprint / triggered by model performance degradation / on client request] Deliver: A deployment script in shell or Python with inline comments at every non-obvious step — written so a developer running it for the first time understands what each block does without asking the author A pre-deployment checklist embedded in the script: 5 automated checks that run before deployment begins — dependency version confirmation, environment variable presence, model artifact checksum, target environment connectivity, and rollback script availability A failure handling section: for each of the 3 most common deployment failures in ML consulting environments, add an error catch that prints a plain-English message explaining what failed, why it likely happened, and the exact command to recover A deployment log output standard: define what the script prints at each stage so a new developer watching the terminal knows whether deployment is progressing normally or silently failing An environment variable reference document: list every variable the script requires, what it controls, what breaks if it is missing, and the safe default value where one exists — formatted as a comment block at the top of the script A rollback procedure: write the rollback script that undoes the deployment if a production bug is detected within 30 minutes of release, with the same inline comment standard as the main deployment script A new developer onboarding addendum: a README section of 200 words that explains how to run the deployment script for the first time, what to do if each pre-deployment check fails, and who to contact if the script produces an error not covered in the failure handling section A production bug prevention audit: review the current deployment process and identify the 3 specific gaps that are causing the recurring production bugs — with the exact line in the script that closes each gap Write the pre-deployment checklist before writing the deployment logic — a script that catches problems before they deploy prevents production bugs that a perfect deployment script cannot fix after the fact.