CapTech IDFIN PROGRESS
Interactive map of Île-de-France technology companies in AI, cybersecurity and embedded systems, built from public data (Insee Sirene). A company is attached to a domain only with cited, dated evidence. Private code.
- ROLE
- Solo · Personal project
- STACK
- TypeScript · Next.js · Supabase · MapLibre · ECharts
- RESULT
- Corpus cleaning cut indexed segments from 2,105 to 1,343 over 45 organisations. Retrieval is benchmarked on 18 queries as positive–unlabelled, so no accuracy or F1 is claimed.
- LINKS
- Demo ↗
- READING TIME
- 1 min read
1
Data
Insee Sirene (company and establishment stocks) and the public company-search API, both under Licence Ouverte; domain names from the INPI company register and text from company websites, used offline only.
2
What I built
Solo
- The data pipeline from the Sirene stocks and the company-search API into Supabase.
- Offline enrichment: company websites crawled within robots.txt, at most one request per second per host, and an evidence corpus cleaned in two passes.
- The interactive map (MapLibre) and the 3D scene of companies and links.
- 455 automated tests.
3
Key choices
- Evidence before labels
- A company is attached to a domain only with a cited, dated source; companies matched by activity code alone stay candidates.
- Similarity finds candidates, it is never shown
- The interface shows an explainable proximity band backed by citations, not a score out of 100.
- No personal data
- Directors and beneficial owners are excluded at the request level.
4
Results
Corpus cleaning cut indexed segments from 2,105 to 1,343 over 45 organisations. Retrieval is benchmarked on 18 queries as positive–unlabelled, so no accuracy or F1 is claimed.
5
Limits
- Activity codes alone over-include: about 40,000 AI candidates in the Paris area before any evidence.
- The scene draws at most 100 companies per domain.
- The proximity score is designed, not implemented.
Questions about this project? → Email me