Case Study · AI Engineering · Quality & Risk

Give a fast-moving product a regression safety net that learns with it.

A central digital-signage management platform had grown faster than its quality process. I built an AI-assisted approach that scouts the application, generates and maintains browser tests, and runs them against staging and production.

SeleniumDockerChromeClaudeAI-assisted QASmoke testsRegression testing
The Problem

A business-critical platform had one developer and a lot of surface area.

The system—internally known as Dash—functioned as the central nervous system for deployed digital-signage fixtures. Devices reported home, analytics flowed through it, content and updates were managed through it, and clients depended on its visible features.

Like many useful internal systems, it had grown through new requests, pivots, partially completed work, maintenance, and changing dependencies. The company needed a way to find breakage before customers did.

Discovery

First learn the test domain, then automate the repeatable parts.

I learned Selenium and the surrounding browser-automation tooling, then used Gemini initially—and later Claude when company constraints changed—to help design an approach that could understand the application well enough to keep a useful regression suite current.

The emphasis was on client-visible smoke tests: the basic behaviors that must keep working even while the underlying application changes.

What I Built

Scout, map, generate, execute, repeat.

The system can inspect a website and source context, scout and map the application, create the objects and instructions needed for browser automation, and run Selenium-driven Chrome sessions inside Docker.

Pull changes
Scout & map
Update tests
Run Chrome
Review failures

Tests can be organized around scaffolding, fixtures, and smoke coverage so the suite has enough structure to remain useful rather than becoming one giant brittle script.

Result

Catch the break before the customer reports it.

Multiple defects found before customer exposure.The regression system surfaced issues in ordinary development and during a significant upgrade involving PHP, security changes, and dependencies across the stack.

The business value is straightforward: reduce the chance that a client becomes the QA department for a live system tied to deployed fixtures and analytics.

What This Demonstrates

AI is useful when it reduces maintenance cost on a changing system.

  • Use AI to maintain context. The application changes, so the test model has to keep learning the new shape.
  • Focus automation on business risk. Client-visible smoke tests matter more than test-count vanity metrics.
  • Build for change. QA tooling has to evolve alongside the application it protects.
  • Keep a human in the loop. AI accelerates mapping and test generation; failures still require engineering judgment.
Work With Me

Have a problem that's still mostly a question?

That's a useful place to start. I work best where discovery, architecture, hands-on building, and enablement overlap.