IITM BS Introduction to DL and GenAI Project (BSDA2001P): Syllabus and Tips

Introduction to DL and GenAI Project (BSDA2001P) is a 2 credit diploma level project in the IITM BS Data Science programme. You need Machine Learning Techniques (BSCS2007) first. Machine Learning Practice (BSCS2008) and the theory course BSDA2001 are corequisites. It is the project half of Option 2 in the Diploma in Data Science.

The project is run by Prof. Balaji Srinivasan and Prof. Ganapathy Krishnamurthi, who also teach BSDA2001.

Handbook check. The handbook agrees on 2 credits and lists the DL and GenAI theory course as the corequisite. It does not mention the MLT prerequisite or the MLP corequisite that the course page shows. Go by the stricter course page, and confirm with support if you are unsure.

What the project work involves

The course page has no week-by-week plan. It lists what you will do instead:

• Hands-on work in PyTorch and Hugging Face.
• Working with large language models and other generative models, for text, images and more.
• Using transformers, diffusion models and fine-tuning methods.
• Taking part in a Kaggle competition. The page talks of earning a Kaggle medal for your profile.
• A technical viva taken by an industry expert, set up to feel like a real interview.

How it is graded

The course page names a Kaggle competition and a viva, but gives no marks split. The handbook's project grading rules (two viva levels with cutoffs) name the MAD 1, MAD 2 and MLP projects, not this one. Do not assume the same rules apply. Ask the course team at the start. The general rules are in how IITM BS projects are graded.

How to plan the term

• Week 1: set up a Kaggle account, a GPU notebook and a Git repository. Read the competition rules fully.
• Early weeks: build a simple baseline and submit it. A working baseline early beats a clever model that never runs.
• Middle weeks: improve one thing at a time. Log every experiment with its score in a small table.
• Before the viva: make sure you can explain every step, from data cleaning to the final model. Practise saying why you chose each model and what you tried that failed.

Academic integrity

This is graded as your own work. Public Kaggle notebooks are fine to read for learning, but copying one and submitting it as yours is not. Follow the competition's rules on teams and code sharing. Do not share your code with batchmates. The viva checks whether you understand what you submitted, and copied work tends to fall apart there.

Common questions

Can I take this project in a later term than BSDA2001?

Yes. A corequisite means BSDA2001 must be done before the project or in the same term, so a later project term fits. You need both parts of Option 2 to finish the diploma. See the Diploma in Data Science.

Do I need a Kaggle medal to pass?

The course page does not say so. It lists a medal as a way to strengthen your profile and does not call it a pass rule. Ask the course team what the passing bar is.

Will this project help my portfolio?

It can, if you document it well. See portfolio projects for IITM BS students for how to present it. The theory side is covered in Introduction to DL and GenAI (BSDA2001).