Back to work
In progressSaaS2026 — present
Stack AI Africa
AI document processing: extraction, chat and a model testing bench
An AI document-processing platform built as microservices: a Laravel API orchestrates three FastAPI services — schema-driven extraction, conversational chat with history, and a playground for comparing models. Extraction is guided by a user-defined schema, which makes the output storable in a database rather than merely readable.
- 3
- AI microservices
- 92
- API routes
Role
Architect & developer — Laravel API, AI microservices, dashboards
Stack
Laravel 12PHP 8.3FastAPIPythonPostgreSQLRedisCloudflare R2MinIONext.js 16ZustandRechartsDocker
Highlights
- Three separate FastAPI microservices — documents, chat, playground — orchestrated by a Laravel API
- Schema-driven extraction: the user defines the expected structure, the service returns typed data
- Playground for model comparison and prompt testing, wired to OpenAI-compatible APIs or a local LM Studio
- Complete API domain: 92 routes, 15 models, 10 authorisation policies, spatie permissions
- Trilingual public landing (English, French, Wolof) plus separate internal and client dashboards