A deceptively small question hid an entire product surface.
The assignment began with a broad challenge: build something similar to BrightSign. My first task was not coding. It was figuring out what BrightSign actually did, how our existing internal signage software worked, and which capabilities mattered for the fixtures we needed to support.
An older internal system, used on previous signage work, provided useful reference points but was not the product we needed. I treated it as source material rather than a constraint.
Learn the product category before choosing the architecture.
I worked through documentation, the existing codebase, device behavior, screen-control requirements, and the realities of running software continuously on dedicated hardware. That discovery clarified the actual system: not simply a video player, but an environment coordinating screens, local content, physical interactions, administration, and device health.
- Studied BrightSign concepts and existing internal signage code.
- Mapped required playback, control, content, and administrative behaviors.
- Evaluated which older components were reusable and which should be replaced.
- Used ChatGPT and Gemini to accelerate unfamiliar-domain research and implementation.
A full signage stack from device to admin interface.
I stood up a new development project around a custom Ubuntu-based signage device. The working system combines a Flutter front end on the device, a React administrative/demo interface, Python-based components, local content-delivery capabilities, and support for independently controlled screens.
A physical button controller can trigger videos, volume changes, and other fixture interactions. That matters because retail signage is physical software: the product only works if buttons, screens, media, operating system, and management tools behave as one system.
Use AI where it shortens diagnosis, not where it creates theater.
AI became part of the engineering and operational workflow rather than a decorative feature. In the development environment, Gemini could connect to a networked device, inspect properties and logs, and help investigate performance or configuration problems.
The useful idea was not “add AI to signage.” It was give the engineering workflow better eyes and a faster feedback loop.
A working end-to-end prototype built for rapid iteration.
The project remains an evolving platform rather than a finished commercial product, so I do not claim deployment metrics that do not yet exist. Its value is as evidence of rapid domain learning, architecture, hardware/software integration, and product decomposition under ambiguity.
Creative technology is still systems engineering.
- Translate an analogy into requirements. “Like BrightSign” is a starting point, not a specification.
- Work across boundaries. Hardware, Linux, UI, media, physical controls, and administration all matter.
- Prototype visibly. Make the system tangible early enough for stakeholders to react to pixels and behavior.
- Use AI pragmatically. Put it where it reduces discovery and debugging time.