This project demonstrates an approach to video content recommendations using machine learning and Generative AI. By analyzing semantic relationships rather than just metadata tags, this technique offers an effective solution for content recommendation.

• Data Analysis: Examines metadata from 150 recent streaming titles to identify content similarities and relationships
• Semantic Vectorization: Transforms content's metadata into dense vector embeddings using a text embedding model
• Dimensional Analysis: Employs Principal Component Analysis (PCA) for dimensionality reduction and K-Means for cluster identification
• Visual Representation: Creates interactive 2D scatter plots to visualize content relationships and clustering patterns
• AI-Powered Insights: Leverages Generative AI to produce natural language descriptions of each content cluster and extract meaningful keywords

This approach enables content platforms to automatically group similar videos based on deep semantic connections rather than superficial categorization, significantly enhancing recommendation accuracy and content discovery.

ollama pull qwen3-embedding:4b
ollama pull gpt-oss:20b

python -m pip install virtualenv --break-system-packages -Uq
python -m venv .venv
source .venv/bin/activate

python -m pip install pip -Uq
python -m pip install -r requirements_ollama.txt -Uq

cp env.txt .env # update file after copying

python -m pip install virtualenv --break-system-packages -Uq
python -m venv .venv
source .venv/bin/activate

python -m pip install pip -Uq
python -m pip install -r requirements_bedrock.txt -Uq