AI-Assisted Marketing Studio
An application that creates marketing drafts from site photographs and requires human approval before export.
Challenge
Turning photographs from field teams into consistent marketing content takes time. Fully automated generation also creates a risk of inaccurate, exaggerated or unapproved material being published.
Approach and architecture
The application creates text and image drafts from photographs. Export is possible only after the user has reviewed and approved the content. Approval is not merely a warning; it is a mandatory step in the workflow. An independent test implementation allows the main workflows to be tested without an external AI service. This keeps testing possible without network access or ongoing external service costs.
My contribution
I developed the product concept and designed the approval boundary, background-processing structure and testing approach.
Outcome and current status
The core workflow prevents export before approval. Product development and user-experience improvements are continuing.
Available materials
This project is not publicly accessible. I can present selected views and technical details in an appropriate professional discussion.
Technologies used
- TypeScript strict
- Turborepo
- Drizzle
- pg-boss
- Zod
- Vitest + PGlite
- Provider-agnostic AI adapter