WAMOS
The WebWriterStudio AI Marketing Operating System — a multi-tenant SaaS that drafts LinkedIn content in a brand’s own voice, then checks every draft for duplication and readability before a human ever sees it. In build.
- Role
- Independent developer
- Context
- WebWriterStudio · Independent product
- Branch
- Independent
- Stack
- Laravel
Illustrative interface composed from the product’s own specification. Not a screenshot of the running application.
The work
The problem
Generic AI output is the easy part; content that sounds like one particular business is not. A marketing team gets drafts that drift off-brand, repeat what was posted a fortnight ago, and arrive with no way to tell the good ones from the filler — so somebody reads all of it anyway, and the tool saves nothing.
What I built
A single-database multi-tenant platform, isolated per workspace by a global query scope, with six roles from super admin down to viewer and a workspace switcher that never asks for a second login. Brand context — company background, services and USPs, audience personas with pain points and buying triggers — is assembled into the prompt at generation time, alongside a local holidays table so seasonal posts need no runtime call out. Generation routes through a provider adapter to a Kie.ai gateway, queued on Redis and Horizon, metered against plan credits. Before a draft surfaces it passes a duplicate check against the workspace’s last hundred posts, by TF-IDF cosine similarity and Levenshtein distance, and a quality pass scoring length, emoji count, CTA presence and Flesch reading ease. Drafts export to PDF, Excel and Word, with published-post metrics logged back by hand.