Streamify ML — Machine Learning with CI/CD

A machine learning system for content recommendation and stream forecasting on a streaming platform, deployed on AWS using Lambda, DynamoDB, and S3.

    ∧＿∧　 ♡      dont give up,
  (｡･ω･｡)つ━☆    every bug you fix
  (　　　 )         makes you stronger!
  | | |          with love & coffee ☕
  (＿)＿)          for lkscc fighter!

Architecture

Component	Description
ETL (AWS Glue / Spark)	Processes raw interaction data from S3 and prepares features
Lambda - Prediction	Predicts the probability that a user will stream a given content
Lambda - Forecasting	Forecasts content stream counts using time-series methods
Flask App	Web frontend that proxies requests to Lambda via API Gateway
Dataset Generator	Generates synthetic streaming platform datasets for development

Environment Variables

API Gateway

PREDICTION_API_URL=<API Gateway URL for prediction>
FORECAST_API_URL=<API Gateway URL for forecasting>
API_GATEWAY_KEY=<API Gateway key>

AWS Credentials

AWS_ACCESS_KEY_ID=<access key>
AWS_SECRET_ACCESS_KEY=<secret access key>
AWS_SESSION_TOKEN=<session token>

Flask Configuration

FLASK_HOST=0.0.0.0
FLASK_PORT=5000
FLASK_DEBUG=true

Project Structure

core/
├── dataset/                  # Synthetic dataset generator
├── ETL/                      # AWS Glue / Spark ETL job
├── lambda/
│   ├── lambda_forecasting/   # Stream forecasting Lambda function
│   └── lambda_prediction/    # Content prediction Lambda function
├── machine_learning/         # Model training notebook
└── newfront/                 # Flask web application

Datasets

File	Description
user_profiles.csv	1,000 user accounts with subscription plan and watch behavior
content_catalog.csv	500 content items (movies, series, music, podcasts)
user_interactions.csv	10,000 interaction events (play, like, complete, skip, add to watchlist)
subscription_history.csv	Subscription records per paying user

Getting Started

1. Configure environment variables as listed above.
2. Upload trained model artifacts to S3 at the path specified by MODEL_KEY / FORECASTING_MODEL_KEY.
3. Ensure DynamoDB tables exist with the names configured in each Lambda's environment variables.
4. Deploy Lambda functions and expose them through API Gateway.
5. Run the Flask application:

cd newfront
pip install -r requirements.txt
python app.py