Welcome to the official repository for CAIC Summer of Tech (CSoT'26) — a beginner-friendly initiative by the Co-curricular and Academic Interactions Council (CAIC), IIT Delhi, designed to give students structured, project-based exposure to multiple development tracks over the summer.
Inspired by DevClub's Summer of Code, CSoT is the launchpad for freshers (and anyone curious) into the wider tech culture at IIT Delhi.
- Project-Based Learning — every week ships hands-on code, not just theory.
- Low Barrier to Entry — execution is intentionally kept accessible so that participation outweighs perfection.
- Inter-IIT Tech Meet Prep — projects are loosely themed around past Inter-IIT problem statements.
- 5–6 Week Tracks — each track runs in weekly modules so you can pace yourself.
- Mentorship — tracks are led by relevant CAIC clubs and societies, with feedback channels open throughout.
| Track | Description | Folder |
|---|---|---|
| ML in Astronomy | Build a Convolutional Neural Network in PyTorch to classify galaxy morphologies (Galaxy Zoo 2 dataset) on Google Colab. | ML-Astronomy/ |
More tracks will land here as they're published.
| Week | Theme | Status |
|---|---|---|
| 1 | Foundations: Tensors, GPUs & the Galaxy Data Pipeline | Available |
| 2 | Baselines & Fully-Connected Networks | Available |
| 3 | CNNs, the Training Loop & Evaluation | Available |
| 4 | Multimodal Lens Finding with CLIP | Available |
| 5 | VLM Audit and the CNN vs CLIP vs VLM Capstone | Available |
Read the track-level overview for the project abstract, dataset details, and the full learning arc.
You don't need to be an expert in either machine learning or astronomy to start. You will, however, get the most out of the track if you have:
- Python basics — variables, loops, functions, lists, dicts, and a passing familiarity with NumPy.
- A Google account — Google Colab is our default execution environment (free GPU runtime included).
- Curiosity about the night sky — useful, not required.
If any of the above is missing, week 1 links out to crash-course resources that get you to "good enough" in a couple of hours.
- Pick a track folder (currently:
ML-Astronomy/). - Open that folder's
README.mdfor the project abstract and weekly roadmap. - Each week lives in its own subfolder (
Week-1/,Week-2/, …) with concept markdowns, external links, and starter/solution notebooks. - Work through the numbered markdowns in order, then complete the week's project task in Colab.
- Compare your work against the solution notebook only after you've genuinely attempted the starter.
CSoT'26 is organised by CAIC, IIT Delhi, with track-specific mentorship from member clubs and societies. Reach out via the official CAIC channels for queries, submission instructions, and community discussion.
Happy hacking, and clear skies. 🛰️