Image recognition service (digits, simple neural network for classification)

Model training in Google Colab (Jupyter Notebook, Data Analysis and Visualization, Deep Learning) using the MNIST Dataset and a simple neural network.

User interaction in a chat (Telegram/Discord, FastAPI, SQL): a) the model is used to classify photos of digits sent by the user in the chat; b) the user confirms the correctness/incorrectness of the answer, and reports the correct answer in case of incorrect classification; c) the user ID, date, image, correctness/incorrectness of the answer, etc. are stored in the database for subsequent analysis/model training.

Admin panel, database access (Flask/Django): a) allows authorized users to work with the accumulated data (for example, save in CSV format); b) provides other applications (e.g., Streamlit) with access to part of this data through its API in JSON format;

A data visualization application (Streamlit) displays statistical information on the accumulated data.

Optional deployment of the developed applications to a third-party service (pythonanywhere, railway, heroku, etc.).