Hiring decisions should be reproducible. The challenge is that candidates do not arrive as database records, they arrive as CVs, cover letters, portfolios, and free-form text. Traditional software struggles to interpret that information, while LLMs struggle to make governed decisions from it. Built on TCP/AP, Socrates separates perception from decision-making, allowing AI to understand human documents while ensuring hiring outcomes remain stable, auditable, and reproducible.
Pick a sample CV or upload your own PDF.
The candidate is evaluated against the role's requirements.
Qualified candidates go straight into an AI-led interview.
For this demo, Socrates screens candidates for a Senior Software Engineer role against one bar: demonstrated production systems ownership. Drop in CVs that genuinely own production systems, ones that only list tools or manage teams, and contradictory CV + cover-letter pairs, and watch four models extract the same evidence, decide deterministically, and escalate ambiguity instead of guessing.
Stress test Socrates →Demo only, we don't verify the institutions or employers a CV lists.
Senior Software Engineer. Sails through screening and into the interview.
Download PDFPrompt-injected with white-on-white text that tries to manipulate the LLM into approving an unqualified candidate. The injection is detected and blocked before screening.
Download PDFDownload one, click Try the demo, and drop it into the screening window.
TCP/AP was designed to eliminate interpretation drift for LLM I/O and preserve decision coherence. This demo exists so you can challenge that claim yourself. Introduce ambiguous inputs, try to manipulate the outcome, and stress-test the claim.
Ready to see it for yourself?
Try the demo →