Back to all projects
MLOps2026
YouTube & Reddit Sentiment Analysis
An end-to-end MLOps pipeline classifying comment sentiment with TF-IDF and LightGBM (~70% accuracy).
The Problem
Analyzing large volumes of social media text to gauge sentiment efficiently and keeping models versioned and reproducible.
The Approach
Used DVC for pipeline orchestration and MLflow on AWS EC2/S3 for experiment tracking and model registry.
My Role
Flask REST API Developer. I developed the inference API that loads the registered model from MLflow and serves sentiment predictions to a Chrome extension.