Creative tooling^†

Performance creative at volume needs tooling, and a lean team rarely gets engineers for it. So I built the layer myself, with AI-assisted development and no engineering resources.

A production-fidelity sandbox of the company's consumer apps let creators and editors shoot real product UI with fully configurable data and no live accounts, which is what made remote creator capture safe. An automated transcript-to-edit pipeline turned interview footage into paper edits. A code-rendered static system produced every variant of a test from one template, including the Catch A/B/C cells. Workflow automations and live dashboards gave the creative org project tracking, work attribution, and bottleneck visibility.

The Catch launch shows what the static system was for. A new product had three plausible stories: an app that pays you for games and tasks, an app that finds money companies already owe you, or both at once. Every cell was the same design with a different argument, across three Meta placements, so the test read the message and nothing else. The winning cell ran 46 percent below blended cost per engagement, and its read went on to steer the product's site and messaging.

The tools are internal, so the sandbox appears here behind bars and the rest is described.