MLOps: MLflow + DVC + AWS to ship a YouTube Sentiment Chrome Extension
What this is: live, end-to-end build of an MLOps workflow that ships a real product: a Chrome extension that analyzes YouTube comments (sentiment, trends, word cloud) through a FastAPI and a trained classifier.
Core flow we will implement
Data to model: TF-IDF features, class weighting, LightGBM baseline, metrics that matter for imbalance.
Experiment tracking: MLflow runs, comparisons, and Model Registry usage.
Pipelines: DVC stages for ingest → clean → featurize → train → evaluate.
Packaging: Docker image for the API, stored in ECR.
CI/CD: GitHub Actions that retrain, register, build, push, and deploy on every merge.
📍 Event Details
📅 Date: Saturday, December 06
🕚 Time: 12:00 PM – 02:00 PM
📌 Location: Online Gmeet - https://meet.google.com/zbn-jkwe-bdx
Audience Software Engineers, Data Scientists, Machine Learning and AI Engineers who want a practical template to move from notebooks to production. No prior MLOps in production required, but you should be comfortable with Python, git, and basic AWS terms.
Format and agenda (120 minutes)
Problem and architecture in 10 minutes
Experiments and tracking in 20 minutes
DVC pipeline and params.yaml in 20 minutes
Registry, tests, and promotion in 20 minutes
Docker and AWS infra (s3, ECR, EC2) in 20 minutes
CI/CD with GitHub Actions in 20 minutes
Q\&A and pitfalls in 10 minutes
Prerequisites for hands-on
Python 3.10+, git, Docker installed.
Optional AWS account if you want to mirror the deploy steps.
Outcome Attendees leave with a working demo, a production-ready pipeline, and a checklist to adapt this pattern to their own projects.