A production-style, event-driven SaaS analytics platform that ingests real-time user activity, generates live operational metrics, predicts traffic spikes using machine learning, and proactively scales microservices using AI-based scaling decisions.

This project demonstrates distributed systems, real-time streaming pipelines, production ML inference, and cloud-native scalability.

• Event-driven microservices architecture with clear separation between ingestion, streaming, processing, and analytics.
• Real-time event ingestion using Apache Kafka (user-events topic) for high-throughput streaming.
• Persistent analytics storage using PostgreSQL, enabling both real-time and historical queries.
• Machine learning traffic forecasting using LSTM models (TensorFlow/Keras) served through an inference API.
• AI-driven auto-scaling controller that scales Kubernetes workloads proactively based on predicted load and system health.
• Live monitoring dashboard built with React, visualizing metrics, predictions, and scaling decisions.
• Observability-first setup using Prometheus + Grafana for metrics monitoring.

• Node.js + Express (API Gateway, Event Producer)
• Apache Kafka (Streaming Layer)
• PostgreSQL (Event + Metrics Storage)

• Analytics Service (real-time aggregation)
• ML Inference Service (FastAPI)
• AI Auto-Scaling Controller

• React + TypeScript (Monitoring Dashboard)

• Docker + Docker Compose (local infra)
• Kubernetes (deployment + scaling)
• Prometheus + Grafana (monitoring)

self-optimizing-saas/ ├── frontend/ # React dashboard ├── services/ # Microservices (producer, consumer, analytics, ML, autoscaler) ├── infra/ # Docker + Kubernetes configs ├── docs/ # Architecture diagram + documentation ├── monitoring/ # Prometheus + Grafana setup ├── scripts/ # Local setup + testing scripts └── docker-compose.yml # Kafka + Postgres local environment