NLP / CLASSIFICATION

Spam Detector — AI Message Classifier

Real-time AI spam detection system. Paste any message and get an instant ham/spam verdict with confidence score. NLTK preprocessing pipeline feeds a scikit-learn classifier served via FastAPI on Render.

Public demo

<100ms — live API on Render

The problem

Client-side message filtering needs a lightweight NLP model that classifies SMS/email spam in under 100ms without sending user data to a heavy cloud LLM — privacy-preserving and fast.

The approach

NLTK tokenisation and lemmatisation clean the input text; TF-IDF vectorisation (5 000-feature vocabulary) converts it to a numerical vector; a Multinomial Naive Bayes classifier with Laplace smoothing outputs ham/spam probabilities. The trained pipeline is serialised with pickle and served from a FastAPI endpoint deployed on Render.

The results

Live REST API at spam-detection-e1hd.onrender.com. Classifies messages in under 100ms with confidence percentage. React glassmorphism frontend with smooth animations deployed on Netlify.

Stack: React 18, Vite, FastAPI, Scikit-learn, NLTK, Python

• Live demo