Nexus AI Recommendation Engine
A high-performance, dual-domain Machine Learning recommendation system that leverages content-based filtering to match and recommend movies and anime with instant inference speeds.
Project Gallery
Key Features
⚡ Lightning-Fast Inference
Reduces standard 6-second API cold-starts to 0.5s using ThreadPool concurrency, with 0.00s latency on cached searches via @lru_cache.
🧬 Dual ML Domains
Independent vector spaces for TMDB 5000 Movies and MyAnimeList datasets with seamless switching.
🏷️ Granular Metadata Scrapers
Dynamically scrapes runtime, studios, genres, and release dates via REST integration with TMDB and Jikan APIs.
🎨 Premium UI/UX
Netflix-inspired glassmorphic overlays, responsive fluid grids, and polished Light/Dark mode mechanics.
🧬 Content-Based Filtering
Advanced NLP-driven matching using CountVectorizer and Cosine Similarity for accurate recommendations.
🐍 Streamlit Prototype
Robust Python-only interactive variant created for rapid data-science demonstrations.
Technical Architecture
Backend Pipeline
FastAPI server with intelligent caching strategies. NLP feature engineering via CountVectorizer, cosine similarity matrix computation, and live metadata scrapers pulling from TMDB and Jikan APIs.
Machine Learning Engine
Content-based filtering using extracted, standardized text features (cast, genres, studio, synopsis). Five-thousand-dimensional vector spaces mapped through cosine similarity for instant neighbor searches.
Frontend Experience
Vanilla JavaScript + Tailwind CSS single-page application. Netflix-inspired glassmorphic cards, smooth animations, responsive grid layout, and dynamic theme switching.
Data Integration
Dual dataset support: TMDB 5000 movies and MyAnimeList entries. Real-time metadata enrichment with granular runtime, studio, genre, and release date scrapers.
The Architecture
📦 nexus-ai-recommendation
├── 📁 api/
│ └── 📄 main.py # FastAPI server (Endpoints, Caching, Live Scrapers)
├── 📁 data/ # Raw CSV Datasets
├── 📁 frontend/
│ └── 📄 index.html # High-Performance Vanilla JS + Tailwind SPA
├── 📁 models/ # Pickled ML models (Similarity Matrices, Pandas Dictionaries)
├── 📁 src/
│ └── 📄 train.py # NLP Feature Engineering & Vectorization logic
├── 📄 streamlit_app.py # Python-only Interactive UI
├── 📄 Kaggle_Notebook.ipynb # Jupyter Notebook export for Kaggle Portfolio
└── 📄 requirements.txt # Dependencies