Machine learning in production
Take models from a notebook to a monitored service, and keep them working once real users depend on them.
- Level
- Advanced
- Language
- English
- Duration
- 30 hours
- Lessons
- 10
Last updated 8 Oct 2026
What you'll learn.
- Explain what changes when a model moves from a notebook to production
- Package a model with a reproducible environment and serve it behind an API
- Prevent train–serve skew with shared feature code
- Choose evaluation metrics that reflect the real cost of errors
- Monitor data drift and model performance after launch
- Roll out new models with shadow and canary releases, and roll back quickly
How this course is taught.
Every module comes to you in four parts, so you can watch, read, discuss and check what you have learned.
- Quadrant I
e-Tutorial
Watch a short video lecture for each module, with full text notes beside it so you can review or translate at your own pace.
- Quadrant II
e-Content
Download or print the reading material for each module, with key terms, case studies and links to further resources.
- Quadrant III
Discussion forum
Ask questions and clear your doubts with other learners. The course team guides the discussion and answers questions.
- Quadrant IV
Assessment
Check your understanding with quizzes, short and long-answer practice and FAQs.
Curriculum.
3 modules · 10 lessons1From notebook to service3 lessons · 57 minReproducible packaging and a prediction API you can deploy.e-Tutorial (video)1e-Content (notes)2
- Note: What changes when a model goes to production15 min
- Video: Packaging a model with a reproducible environment20 min
- Note: Serving predictions behind an API22 min
2Data and model quality3 lessons · 48 minKeep training and serving consistent, and measure what matters.e-Tutorial (video)1e-Content (notes)1Assessment1
- Note: Train–serve skew and feature pipelines20 min
- Video: Evaluating models beyond accuracy18 min
- Quiz: Check your understanding: production ML10 min
3Operating models4 lessons · 4 h 55 minMonitoring, safe releases and deciding when to retrain.e-Tutorial (video)1e-Content (notes)2Assessment1
- Note: Monitoring drift and performance22 min
- Note: Safe roll-outs: shadow, canary and rollback18 min
- Video: Retraining policies15 min
- Assignment: Project: a model service with a monitoring plan4 h
Before you start.
- Experience training models with scikit-learn or a similar library
- Working knowledge of Python and the command line
- Basic understanding of HTTP APIs
About this course.
A model with a good validation score is the start of the work, not the end. In production it has to receive data it has never seen, answer within a time budget, and keep being right as the world changes. This programme covers the engineering that makes that possible.
Who it is for
Data scientists who have trained models and want to ship them, and software or data engineers who are being asked to run machine learning systems.
What you will learn
- Packaging a model so training and serving behave identically.
- Serving predictions behind an API with input validation.
- Choosing evaluation metrics that match the business cost of mistakes.
- Monitoring for drift and rolling out new models safely.
Final project
You will build a small prediction service and write the monitoring and roll-out plan you would use in a real team. Your instructor reviews both.
Your instructor.
Arjun designs cloud platforms and the security and operations practices around them, and teaches cloud, machine learning operations and security for teams.
Learner reviews.
No reviews yetNo reviews yet. Learners can review a course after completing half of it.
Questions about enrolling.
How long can I access this course?
You can access the course for 365 days from the day you enrol.
Is my payment secure?
Yes. Your card, UPI or bank details go straight to Razorpay. Walnut Data Tech never sees or stores them.
When do I get access?
Immediately after the payment is confirmed. You'll also get a receipt by email.
What if the payment fails but money is deducted?
This is usually a delay at the bank. If the payment is confirmed, you're enrolled automatically. If not, the bank reverses the amount, normally within 5–7 working days. Contact support with your receipt number if you need help.
Can I get a refund?
Yes, if you ask within 7 days of purchase and have completed less than 20% of the course. Raise a ticket from the help desk with your receipt number.