- Cloud
- Fintech
- AI
PayTrack Africa
Multitenant serverless invoicing for Ghanaian SMEs, with AI collections
Fintech SaaS for small and medium businesses in Ghana
Project overview
PayTrack Africa addresses the everyday work of invoicing and following up on overdue payments for Ghanaian small businesses. Rather than treating collections as a separate task, the product brings invoice tracking and reminder drafting together. Gemini assists with the wording of payment reminders, helping owners prepare a follow-up without starting from a blank message each time.
The central engineering concern is tenancy: several businesses share the same backend, but their data must remain separate. Cognito handles authentication, while the application scopes access to each account's data in DynamoDB. Lambda and API Gateway provide the serverless request path, and Terraform makes the infrastructure reproducible. The AI integration supports the workflow while identity and data access remain explicit backend responsibilities.
The problem to solve
Small businesses often track invoices in spreadsheets, which makes it hard to see who owes what and easy to let late payments drift. A shared multi-tenant service has to keep every business isolated while staying cheap enough to run for many small accounts.
Inside the build
The implementation, from application behaviour to the infrastructure underneath it.
- Model tenancy so one account can never read another account's data
- Authenticate and scope users with Cognito, without running an identity service by hand
- Keep the backend serverless so idle tenants cost nothing
- Generate context-aware payment reminders with the Gemini API
- Define the stack in Terraform so environments are reproducible
What shipped
A multitenant invoicing service that isolates tenants at the data layer, scales to zero when unused, and turns collections from a manual chore into an AI-assisted one.
Built with
- AWS Lambda
- DynamoDB
- Cognito
- API Gateway
- Terraform
- Python
- Gemini API
Want to talk through how this fits a system you are building?