Back to projects

HSD-system

Role: Backend Software Engineer

HSD-system

Backend of a multirepo hatespeech detection system with OCR and ML models

The backend of the Hate Speech Detector (HSD) project.

For this project we introduced architecture and infrastructure diagrams to the documentation. This was also the first project we did following SCRUM principles.



This project contains two flows. There is the OCR part which reads text from screen, sends the text over websockets to the backend flask server. The other route comes from a seperate webserver that hits the REST API, which is on the same flask server.

Both routes use the same machine learning pipeline and send the result back.



The machine learning pipelines consists of a logistic regression classifier to determine if text has hatespeech or not, then it is send to a RoBERTa classifier model that flags the kind of hatespeech it contains.



The project uses MVC structure.