Peter Foster
27921d625f
- Sort AI candidates by fuzzy match score before taking top 10 This fixes Royal Bank of Scotland matching (was getting arbitrary candidates from Dictionary, now gets most relevant) - Add historical employer recognition (Foster Wheeler, Glaxo, etc.) - Add public sector employer recognition (NHS, councils, etc.) - Add charity/non-profit recognition - Add company division pattern recognition - Improve AI matcher prompt with explicit examples - Add partial company number matching for truncated AI responses - Lower AI confidence threshold to 30% (fuzzy validation as backup) - Add whole-word boundary matching for subsidiary indicators Fixes "SCOTLAND" incorrectly matching "land" pattern - Add 100+ historical polytechnic → university name mappings - Add post-1992 universities and Welsh institutions Results: Employer verification improved from 71% to 85% 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
