IITM BS Machine Learning Operations (MLOps) (BSDA5014): Syllabus and Tips

Machine Learning Operations (BSDA5014) teaches you how to take a machine learning model out of a notebook and keep it running for real users. You learn data pipelines, versioning, deployment, monitoring, security and governance, with a final part on LLMs. It is a 4 credit degree level elective, marked L5 in the handbook. The course page lists Programming in Python (BSCS1002) and Machine Learning Practice (BSCS2008) as prerequisites.

The instructor on the course page is Rangarajan Vasudevan, co-founder and Chief Data Officer of Lentra.ai. He also teaches Introduction to Big Data.

Where the course page and handbook differ

• The course page lists two prerequisites. The handbook's degree level table (updated 18 March 2026) leaves the prerequisite column empty.
• The table calls the course "ML Ops" and tags it BP. The course page leaves the course type blank.

This gap should not block you. Python is a foundation course and Machine Learning Practice is a Diploma in Data Science course. You need both diplomas to reach the degree level, so you will have passed both anyway.

What you learn in MLOps

The course page lists 12 weeks. Here is the plan in simple terms.

• Weeks 1 to 3: the basics and your data. What MLOps is, why models are hard to run in production, and how it differs from normal software work and DevOps. Then data pipelines and how to automate them, data versioning with DVC, and feature stores with Feast.
• Weeks 4 to 6: from code to a running model. Git and automated testing for models (CI/CD), tracking training runs and tuning with MLflow, small models versus large ones such as LLMs, and ways to deploy and serve a model: containers, cloud, edge and on-premises.
• Weeks 7 to 9: keeping it healthy and safe. Watching a model after it goes live, logging and error tracking, scaling, security for ML systems (MLSecOps), explainability, ethics and tracking bias.
• Weeks 10 to 12: LLMs and advanced topics. Versioning base models and fine-tuned versions, the balance between accuracy, speed and cost, observability and governance for LLMs, then federated learning and inference on edge devices.

How MLOps is assessed

The course page does not print its own assessment pattern. It points to the standard structure on the programme's Academics page. Read the grading document for your term before you plan your weeks.

Where MLOps counts

• Minor in Cloud Computing for AI. MLOps is one of its three courses, with Deep Learning Practice and Introduction to Big Data. In the current handbook list it is in no other minor. See the Cloud Computing for AI minor.
• BS level stream rule. The BS level needs 2 programming stream courses at level 4 or higher. MLOps is an L5 course tagged BP, and BP appears to mean the programming stream. The handbook does not define BP, so confirm with support. See the BS degree level.
• Any elective slot. It also fills 4 elective credits at the BSc or BS level.

The handbook says L5 courses are more complex than Degree and L4 courses, but the credits and fee are the same as L4. The March 2026 table marks MLOps as offered in May 2026, September 2026 and January 2027, subject to enough registrations.

Who finds it hard and how to prepare

MLOps is light on maths and heavy on tools. The course page promises hands-on work with MLOps tools, so expect to run commands, not just watch videos. If you have only ever trained models inside Colab, the setup work can feel slow. The hard part is the number of tools, not the depth of any one.

• Before the term, set up Python on your own laptop. Learn to make a virtual environment and install packages without errors.
• Practise basic Git: clone, branch, commit, push and open a pull request.
• Take one model from your MLP work. Save it, load it in a small script and serve it through a tiny web app. Your MAD 1 experience with web apps helps here.
• Run one ready-made Docker image so containers are not new in week 6.

What to take before and after

• Before: Programming in Python and Machine Learning Practice, as the course page lists.
• With or after: Deep Learning Practice and Introduction to Big Data complete the minor. Deep Learning Practice needs Deep Learning (BSCS3004) first.

Common questions

Do I need a cloud account or cloud course before MLOps?

No cloud course is listed as a prerequisite. Cloud deployment comes up in week 6 as one of several options. Comfort with the command line will help you more than any one cloud platform.

Is MLOps harder than other degree electives?

The handbook marks it L5 and says L5 courses are more complex than Degree and L4 courses. How hard it feels depends mostly on how much you have used Git, terminals and Python packages before.

Which books does the course suggest?

The page lists four O'Reilly books. They include Building Machine Learning Powered Applications by Emmanuel Ameisen and Designing Data-Intensive Applications by Martin Kleppmann. The page calls them suggested books.