This is a simple end-to-end implementation of Collaborative Filtering algorithm (Nearest Neighbors to be precise). Users can retrieve the output of its algorithm from given input area, no need to run in the outside (e.g. Jupyter Notebook).

For developing the model (later I exported to .joblib file), I use Kaggle Notebook, which I think it is plenty and great because I don't need to mount/upload the dataset, I just need to find the suitable dataset in Kaggle and made new notebook.

Then, for the backend, I use FastAPI, helped with Gemini to write the logic (since I'm too lazy to do that), and then wrapped it with Docker (I ask Gemini to do it again) and I deployed it in HuggingFace spaces. For free tier and small model, it's plenty and very easy to set up.

Last, for the frontend, I use Next.js, my current favorite framework. Again, I ask Gemini to write the API route and the logic, the rest is my work (I like doing the frontend).

Kaggle Notebook : Song Recommendation - Collaborative Filtering

HuggingFace Space : Spotify Recommender API

If you want to know more about me, you can visit my personal page through : dhimashdr.vercel.app