§07·2025Live API
Sentiment Analysis API
Classical ML end-to-end: training notebook → containerised API → public cloud deploy.
Summary
TF-IDF + Logistic Regression on IMDb (~89% accuracy), wrapped in FastAPI, containerised, and deployed to Google Cloud Run with a Streamlit frontend.
The problem
Most ML projects die in the notebook. This one shows the boring-but-essential half: serving, packaging, deploying.
Approach
- 01Train classical model, persist with joblib.
- 02FastAPI service with Pydantic validation.
- 03Docker image deployed to Cloud Run, frontend on Streamlit Cloud.
Highlights
- ·~89% accuracy on IMDb reviews
- ·Auto-generated OpenAPI docs
- ·Cloud Run autoscaling
- ·Public demo + public API