The Problem

Running Playwright tests at scale is slow. Sequential execution burns hours. CI pipelines bottleneck. Teams wait instead of ship. TestGenX needed an execution engine that could run thousands of tests in parallel, spin up infrastructure on demand, and tear it down when done. Fast. Reliable. Cost-efficient. I built that engine.

What I Built

A parallel test execution system that prepares base images for Playwright, spins up virtual machines on Azure, distributes test cases intelligently, runs them simultaneously, and collects results. All orchestrated through a microservices architecture. Owned it end to end. This is forward deployed engineering in action - architecting the system, writing every service, deploying to production, and iterating based on real usage. Platform engineering ownership from start to finish. Learn more about my platform engineering approach and see how this project connects to my other backend systems work.

Architecture

Orchestrator
The brain. Receives test execution requests, calculates optimal VM distribution, queues jobs, and monitors progress. Built with Node.js. Image Builder
Prepares Docker base images pre-configured for Playwright. Browser binaries baked in. Dependencies locked. Startup time minimized. VM Provisioner
Spins up Azure VMs on demand. Scales from 1 to 200 in minutes. Tears down automatically when tests complete. Pay only for what you use. Test Distributor
Splits test suites intelligently across available VMs. Balances load. Handles retries. Collects results in real-time. Results Aggregator
Consolidates results from all VMs into a single report. Pass/fail counts, execution times, failure screenshots, trace files.

The Stack

Compute: Azure VMs with auto-scaling
Containers: Docker, Kubernetes, Helm
Backend: Node.js microservices
Data: MongoDB for job state, Redis for queues
CI/CD: GitHub Actions
Framework: Playwright

Impact

200 parallel VMs supported simultaneously.
7 hours saved per week on test execution.
Internal tool powering client service contracts.
Zero public access - competitive advantage kept in-house.

How It Worked

• Engineer triggers test run via API or CLI
• Orchestrator calculates VM count needed
• Provisioner spins up Azure VMs with base image
• Distributor assigns test files to each VM
• VMs execute Playwright tests in parallel
• Results stream back in real-time
• Aggregator compiles final report
• VMs terminate automatically

From trigger to teardown - fully automated.

AI-Ready Architecture

The system is designed for the future of testing: Intelligent Test Selection
Architecture supports AI-driven test prioritization - running high-risk tests first based on code change analysis. Smart Distribution
Test allocation algorithm optimized for parallel execution, with hooks for ML-based load prediction. Extensible for AI Test Generation
Modular design allows integration with AI-assisted test generation tools, enabling automated test creation from code changes.

Technologies

Node.js · Docker · Kubernetes · Helm · Azure VMs · MongoDB · Redis · GitHub Actions · Playwright · AI-Ready Architecture