← All work
Neuro-symbolic AI for credit
Research on AI that combines learned models with explicit rules and knowledge graphs, applied to credit and compliance.
Context
Independent research, 2025 to 2026.
My role
Researcher and author.
The problem
Credit and compliance decisions must be explained to regulators and investment committees. Language models alone explain poorly.
What I did
- Technical overview for non-specialists
- Comparisons of companies and approaches
- Briefings for companies and investors
What I took from it
Decision quality, data quality, and workflow quality are separate problems. Different AI approaches fit each one.
Updated September 2026