AI Development & Categorization Logic
We used OpenAI’s API combined with custom logic in Python to build a lightweight machine learning layer. It classified transactions based on context, merchant data, and semantic cues.
For example, purchases from SaaS vendors were labeled as “Tools,” while payroll-related transactions were grouped as “Employee Costs.” The AI system also learned from manual corrections, improving accuracy over time.
Frontend: Dashboard in React
Our frontend team built an intuitive React-based interface. Users could drag-and-drop CSVs, edit categories inline, view summaries in real time, and export structured reports.
The dashboard also included visual breakdowns by category, month-to-month comparisons, and a quick view of top spending areas, helping the founding team stay lean and informed.
Backend & Integration
The backend was powered by FastAPI and Python. We applied our expertise in artificial intelligence software development to ensure the system was modular, secure, and ready for real-time AI-powered processing. It handled uploads, triggered AI classification, and stored results securely. We implemented a modular system to make future integration with APIs (like Xero or Revolut) simple when needed.