Welcome to REMLA25 Team 21's project! 🚀
This organization contains a complete Development Operations (DevOps) and Machine Learning Operations (MLOps) project that demonstrates sentiment analysis using modern software engineering practices, containerization, and cloud-native technologies.
Our sentiment analysis application performs real-time sentiment classification on user feedback using a machine learning model. The project showcases a complete MLOps pipeline with microservices architecture, monitoring, testing, and deployment automation.
🌐 Live Demo: Access our application through the deployment instructions below!
The project is organized into 6 specialized repositories, each serving a specific purpose in our DevOps and MLOps pipeline:
| Repository | Purpose | Tech Stack |
|---|---|---|
| operation | 🚀 Main entry point - Orchestration & deployment | Kubernetes, Helm, Docker Compose, Istio |
| app | 🖥️ Frontend UI & backend service | Flask, HTML/CSS/JS, Prometheus metrics |
| model-service | 🤖 ML model serving API | Flask, REST API |
| model-training | 🧠 ML training pipeline | Python, scikit-learn, DVC |
| Repository | Purpose | Usage |
|---|---|---|
| lib-ml | 🔧 Data preprocessing utilities | Shared across training & serving |
| lib-version | 📋 Version management | System versioning & metadata |
- Sentiment Classification: Restaurant review sentiment analysis
- Model Versioning: DVC-based data and model versioning
- A/B Testing: Multiple model variants with traffic splitting
- ML Testing: Comprehensive test suite for model quality
- Containerization: Docker containers for all services
- Orchestration: Kubernetes deployment with Helm charts
- Service Mesh: Istio for traffic management and observability
- Monitoring: Prometheus metrics with Grafana dashboards
- Metrics Collection: Custom Prometheus metrics for ML operations
- Session Tracking: User behavior and satisfaction monitoring
- Performance Monitoring: Response time and throughput metrics
- Sticky Sessions: User-consistent A/B testing experience
- CI/CD: Automated testing and deployment pipelines
- Code Quality: Pylint, Flake8, and Bandit for code analysis
- Security Scanning: Automated vulnerability detection
- Documentation: Comprehensive API documentation with Swagger
Our application includes comprehensive monitoring:
- Business Metrics: Sentiment prediction distribution, user satisfaction
- Technical Metrics: Response times, error rates, resource usage
- User Metrics: Session duration, rating distribution
- Model Metrics: Prediction accuracy, model usage by variant
Access monitoring dashboards through the Kubernetes deployment options above.
This project was developed as part of the REMLA (Release Engineering for Machine Learning Applications) course. Each repository contains detailed contribution guidelines and development setup instructions.
This project is developed for educational purposes as part of REMLA25 coursework.
🚀 Ready to get started? Head to the operation repository for detailed deployment instructions!
For questions or issues, please check the individual repository README files or create an issue in the relevant repository.