All work
2025 – PresentTech Lead → Project ManagerClosed-source
Theseus FinSync AI
Pulls .xlsx data from SharePoint via Microsoft Graph and turns it into financial recommendations — one analysis pipeline, two surfaces (form + chat), no re-uploads.
PythonFastAPIReact 18PostgreSQLAzure OpenAIMicrosoft GraphSharePointDockerAzure Web Apps
2
User surfaces
SharePoint
Data source
2–3
Team size
Summary
The source-of-truth Excel files already lived in SharePoint; making users re-upload them would have introduced data drift and broken the client's compliance model. Theseus reads them where they live (Microsoft Graph, app-only OAuth), normalises real-world Excel mess (merged cells, blank rows, text-stored numbers), and feeds Azure OpenAI through one analysis pipeline shared by two user surfaces — a structured recommendation form and a chatbot for ad-hoc follow-up. Docker → ACR → Azure Web App on every push to development.
Highlights
- Microsoft Graph + SharePoint as data source — data drift eliminated, compliance intact
- Two surfaces (form + chat) sharing one analysis pipeline, no duplicated LLM logic
- Zero-touch CI/CD: push to development → Docker → ACR → Azure Web App
This project is closed-source (built for a Kcube AI client). I'm happy to walk through the architecture, trade-offs, and code on request.