🎌 Anime Recommendation System (MLOps Project)

📌 Project Overview

This repository presents a Hybrid Anime Recommendation System built with state-of-the-art Machine Learning and end-to-end MLOps practices.
The system leverages Collaborative Filtering and Content-Based Filtering to deliver personalized anime recommendations to users.

Unlike traditional ML projects, this solution is fully production-ready, integrating automation, scalability, and monitoring. It combines ML pipelines with cloud-native deployment to demonstrate real-world engineering skills.

🔑 Highlights:

• ✅ Hybrid recommender system (Collaborative + Content-Based)
• ✅ MLOps-driven workflow (DVC, Comet ML, Jenkins, Docker, Kubernetes)
• ✅ Cloud-native architecture deployed on Google Kubernetes Engine (GKE)
• ✅ Scalable, reproducible, and automated end-to-end ML lifecycle

---

🚀 Features

• 🎯 Personalized Recommendations → Top-10 anime suggestions per user
• 🔄 Hybrid Recommendation Engine → Combines user behavior + anime metadata
• 📦 MLOps Integration → Data versioning, experiment tracking, CI/CD pipelines
• ☁️ Cloud Deployment → Google Cloud Storage + Kubernetes cluster (GKE)
• 📊 Experiment Transparency → Metrics & embeddings logged via Comet ML
• ⚡ Scalability & Reliability → Autoscaling pods, load balancing, and CI/CD

---

🛠️ Tech Stack

🔹 Core Development

• Python 3.8
• Flask (backend web app)
• TensorFlow / Keras (deep learning models)
• Pandas, NumPy (data wrangling)

🔹 MLOps & DevOps

• DVC → Dataset & model versioning
• Comet ML → Experiment tracking & visualization
• Docker → Containerization for portability
• Kubernetes (GKE) → Orchestration & scaling
• Jenkins → CI/CD automation
• Google Cloud Storage (GCS) → Artifact & dataset storage

---

⚙️ System Architecture

flowchart TD
    A[User Input: User ID] --> B[Flask Web Interface]
    B --> C[Hybrid Recommendation Pipeline]
    C --> D[Collaborative Filtering: User-Anime Interactions]
    C --> E[Content-Based Filtering: Metadata & Synopsis]
    D & E --> F[Score Fusion & Ranking]
    F --> G[Top-10 Anime Recommendations]
    G --> H[Results Rendered on Web Page]

Loading

---

🤖 Machine Learning Approach

1️⃣ Collaborative Filtering

• Learns latent user–anime factors from rating matrix
• Embedding model (RecommenderNet) with:
  • user_id → user_embedding
  • anime_id → anime_embedding
• Optimized with binary crossentropy for recommendation accuracy

2️⃣ Content-Based Filtering

• Leverages metadata features: genres, synopsis, anime type, ratings
• Computes semantic similarity between anime via embeddings

3️⃣ Hybrid Model

• Weighted combination of collaborative and content-based predictions
• Balances personalization with content diversity
• Produces ranked top-10 recommendations per user

---

🔧 MLOps Workflow

📈 CI/CD with Jenkins + Docker + Kubernetes

flowchart LR
    A[GitHub Commit] --> B[Jenkins CI/CD Pipeline]
    B --> C[Install Dependencies & Run Tests]
    C --> D[DVC Pull Data & Models from GCS]
    D --> E[Build Docker Image]
    E --> F[Push Docker Image to Google Container Registry]
    F --> G[Deploy to Kubernetes: GKE]
    G --> H[LoadBalancer Service → Exposes Flask Web App]

Loading

Pipeline Stages:

1. Code Integration → GitHub → Jenkins
2. Data Retrieval → DVC pulls datasets & artifacts from GCS
3. Containerization → Build & push Docker image to GCR
4. Deployment → Apply Kubernetes manifests (deployment.yaml)
5. Scaling → Kubernetes HPA ensures high availability

---

📊 Experiment Tracking (Comet ML)

Every experiment is logged to Comet ML Dashboard, including:

• 📉 Training & validation losses
• 🧩 User & anime embeddings
• 🔧 Hyperparameters & metrics
• 📦 Model checkpoints

This ensures comparability, reproducibility, and optimization of models.

---

📂 Repository Structure

.
├── application.py          # Flask web app
├── pipeline/               # ML training & prediction pipelines
├── src/                    # Core ML code (data processing, models)
├── config/                 # Config files & paths
├── artifacts/              # DVC-tracked datasets & models
├── static/                 # CSS, JS & frontend assets
├── templates/              # HTML templates (Flask)
├── deployment.yaml         # Kubernetes spec
├── Dockerfile              # Docker build config
├── Jenkinsfile             # CI/CD pipeline
├── requirements.txt        # Python dependencies
└── setup.py                # Package setup

---

👨‍💻 Author

SAMI-CODEAI

---