Data Science & AIApplied AIAdvancedFeaturedTrending

Machine Learning in Production

From notebook to deployed model — drift, monitoring, and the parts nobody teaches.

4.9 (156) 48 lessons · 6 modules 16h890 studentsLanguage: English

Taught by

Meera Krishnan

Principal Data Scientist at Swiggy · Bengaluru

What you’ll learn

  • Serving a trained model behind a real API
  • Detecting data drift and concept drift in production
  • A/B testing models safely
  • Retraining pipelines and CI for ML
  • The monitoring stack that catches problems before users do

About this course

A course for data scientists who can train models but want to ship them. We cover model serving, drift detection, monitoring, and retraining pipelines — the unglamorous engineering work that decides whether your model helps anyone. Taught by a principal data scientist running models in production.

Curriculum

6 modules · 48 lessons · 16h

1

From notebook to service

2 lessons

Serving a model is not the same as training one.

  • Model serialization and loadingPreview20 min
  • FastAPI for model serving25 min
2

Monitoring and drift

2 lessons

Knowing when your model is wrong.

  • Data drift detection22 min
  • Evidently for model monitoring18 min

Requirements

  • Python and scikit-learn fluency
  • Some experience training ML models
  • Basic Docker comfort

Who this course is for

  • Data scientists moving to ML engineering
  • Backend engineers deploying ML models
  • Anyone whose model works in the notebook but breaks in production

Your tutor

Meera Krishnan

Principal Data Scientist at Swiggy · Bengaluru

4.9 (156)1.8k students2 courses

ML practitioner focused on recommendation systems and forecasting. Built Swiggy’s delivery-time prediction model and the menu ranking system. Teaches ML with a strong emphasis on production realities — drift, monitoring, and the boring parts that actually matter.

Student reviews

4.9

156 ratings

5
78%
4
16%
3
4%
2
1%
1
1%

Anil George

5 Sept 2024

31 found this helpful

This should be a mandatory course

Every DS bootcamp teaches you to train models. Nobody teaches you what happens after. Meera’s course fills that gap and then some.

Kavya N

22 Aug 2024

24 found this helpful

Saved my team a painful incident

We set up the drift monitoring from this course and caught a feature distribution shift two weeks before it would have degraded our recommendations.

Frequently asked questions

Do I need deep learning experience?
No. The course uses scikit-learn and XGBoost examples. The production patterns apply to any model.
Is this about MLOps platforms like MLflow?
We use MLflow for experiment tracking, but the focus is on the patterns, not the tool.