Dacosta Workstation
AI-Assisted Election Results Collation
An owner-built mobile and backend prototype for extracting and checking reported figures from photographed statement-of-poll forms.
- Status
- Prototype
- Categories
- Mobile · Backend · AI / Computer Vision
- Role
- Owner and Sole Full-Stack Developer
- Context
- Owned and developed by Dacosta Agyei as an independent proof of concept.
Problem
The prototype explores how photographed statement-of-poll forms could be converted into structured reported figures with arithmetic checks before submission to a collation workflow.
Work performed
Built the Flutter capture flow, FastAPI service, vision-assisted figure extraction, arithmetic checks, and structured submission model.
Selected features
- Mobile capture of photographed statement-of-poll forms
- Vision-assisted extraction of reported figures
- Arithmetic checks
- Structured submission to a central collation workflow
Engineering decisions
- Kept photographed forms as the human-readable source while transforming extracted figures into a structured submission.
- Preserved an explicit prototype boundary: the application was never deployed and was never used in an election.
Outcome
Prototype only; it was never deployed and was never used in an election.
Verified stack
Flutter · FastAPI · OpenAI vision model
No election body, live election, or production deployment is associated with this prototype.