Course overview
Start hereAI for Machine Learning Engineers
START HERE

Priya's Thursday, two ways

8 lessons · 26 prompts

A day in the life of a Machine Learning Engineer: what changes with these prompts.

Track progress as a member

Priya, a machine learning engineer at a small health analytics company.

Priya starts Thursday with a messy appointment table. The table has age, visit history, zip code, and a no-show label. Some rows are missing visit counts, and a cancellation date column could leak the answer. She opens ChatGPT and uses the data cleaning prompt to list missing values, ask for three handling options, and explain which one fits a first model.

Next she moves to feature engineering. She asks Claude to turn visit history into recent and long-term counts, then create a time-based train, validation, and test split. The prompt reminds her to check that no future visit appears in the training rows. She runs the code on a small sample and fixes one date format issue.

After lunch, Priya uses the evaluation prompt in Gemini. She asks for precision, recall, and a list of false negatives with patient visit patterns. The output helps her see that the model misses people with very few past visits. She writes a short note for the team and plans a new feature.

Before she leaves, she uses the deployment prompt in ChatGPT to draft a FastAPI endpoint, a Dockerfile, and an input schema. She reviews the generated code, changes the model path, and sends it to a teammate for review. With the time saved, she leaves at 5:30, walks home, and cooks a proper dinner instead of ordering takeout.

Before

  • Messy tables eat the morning
  • Loss curves take all afternoon
  • Deployment notes live in my head
  • New research waits on the weekend

After this course

  • Clean data ready before standup
  • Evaluation notes explain each miss
  • Serving draft reviewed by lunch
  • Home by 5:30 for a walk

What you'll learn

  • Data cleaning: Turn raw tables into a clean dataset with clear choices about missing values and outliers.
  • Feature splits: Create useful features and reproducible train, validation, and test splits without leakage.
  • Model design: Choose a model family, loss, and output shape, then turn a paper idea into runnable code.
  • Training runs: Diagnose training failures, read loss curves, and plan a sensible hyperparameter search.
  • Error analysis: Pick the right metrics and inspect where your model fails on real examples.
  • Serving models: Package a model into an API, container, and optimized format for production.
  • Drift monitoring: Detect drift, log live behavior, and reduce inference latency over time.
  • Clear communication: Document and explain models clearly, and absorb new research faster.

How this course works

  1. 8 lessonsOne task of your job each, from data preprocessing & cleaning to communication, docs & learning.
  2. Ready-to-paste promptsCopy, fill in the parts in {{brackets}}, paste into ChatGPT, Claude or Gemini.
  3. Tick and completeTick the prompts you tried and mark each lesson complete.
  4. Get certifiedFinish and keep the prompts as your own library.