In this repo, I decided to go beyond notebooks and implement a dockerized Api of the NewsCategorization project described here while following mlops and software engineering best practices.
The project involves
- packaging the sklearn model development process following object-oriented programming best practices
- serving the model using Fastapi
- reproduciability with docker
Further steps would have been
- integrating a CI/CD pipeline with github actions for continuous deployment of the api endpoint on heroku
- building a streamlit frontend that consumes the api. I did not include those for reasons related to already having a gradio frontend that showcases the model capabilities and cost.