Ginqel OffrePersonal project
Structure tender processing
A service-oriented architecture to collect, structure and enrich tender data.
Personal project built for experimentation and skills development.
Visit website — Ginqel Offre
Business problem
Tender data arrives in heterogeneous formats. Using it requires collection, normalization and information extraction before business processing.
My role
Personal project: architecture and development of ingestion, document-processing and business-data services.
How it works
Python and FastAPI handle ingestion. LangChain and Ollama support structured extraction. Spring Boot and Spring Batch provide APIs and batch processing, while PostgreSQL stores the data.
An architecture decision
Separating AI processing, batch jobs and business services isolates their responsibilities and supports independent Docker containers.
Presented features
Document ingestion · Structured extraction · Batch processing · Business API
Security & human validation
AI processing, batch jobs and business services have separate responsibilities. LLM extraction needs evaluation on representative documents before use.
Limitations and project status
LLM extraction needs evaluation on representative documents. Source quality, formats and extraction errors determine output reliability.
Ginqel Offre
Want to discuss the architecture or the project?
Questions about the technical choices, architecture or progress of this personal project? Get in touch.
Discuss this project