This project was born out of pure frustration.

I built multiple ML/DL models, tried different datasets, tuned hyperparameters, and even leaned heavily on ChatGPT. Still, my models kept failing — wrong outputs, poor accuracy, silent bugs. The worst part? 👉 I didn’t know why.

I realized the real problem wasn’t the model — it was lack of visibility into what was going wrong and where. That’s when I decided to build something that helps understand models, not just run them. This project taught me lessons that tutorials never do:

How small preprocessing mistakes can completely break a model

Why evaluation metrics matter more than accuracy

How to debug ML pipelines step by step instead of blindly retraining

That using AI tools is helpful — but thinking like an engineer is critical

I also learned how to convert confusion into clarity.

Built With

  • and-google-gemini-api-to-analyze
  • api
  • debug
  • fastapi
  • gemini
  • json
  • pydantic
  • python
  • streamlit
  • tensorflow
Share this project:

Updates