Situation
Supplier quotations arrive in inconsistent formats, with missing fields and commercial edge cases that make free-form LLM ranking risky.
Procurement intelligence
QuoteSense
A quotation-analysis workflow that deliberately separates probabilistic document extraction from deterministic business logic. LLM output is normalized into typed structures before scoring, risk checks, anomaly detection, and comparative reporting.
Project story
Task
Create a controlled quotation pipeline that separates probabilistic extraction from deterministic procurement decisions.
Action
- Normalized PDF, DOCX, TXT, and XLSX content into typed quotation structures.
- Validated extracted data before calculating completeness, cost scores, comparisons, and anomalies in Python.
- Kept evidence snippets/pages/sheets alongside extracted fields so recommendations could be reviewed instead of treated as black-box output.
Result
- Produced explainable quotation comparisons where arithmetic remains deterministic and qualitative reasoning stays grounded in validated data.
- Shipped the workflow as a Streamlit dashboard with a FastAPI service and deployment-friendly container stack.
Interactive demo · BYOK
Try QuoteSense live
Bring your own provider key to create a short-lived session for this project. The portfolio does not store the credential.
Used once to create the temporary portfolio session. It is not stored by the portfolio.
One session · reusable across project pages · provider key never stored