First pass
The sweet spot
Begin with small and midsized accounts. Their audiences are a free, noisy signal that an existing algorithm found something—but not proof, and not fame large enough to make discovery redundant.
Use AI to find talented, intelligent Americans who may be overlooked or lack conventional credentials.
Software, writing, and independent projects can reveal ability that a job title or résumé does not. Underplaced finds promising evidence, asks a human to review it, and invites candidates to take a difficult assessment before connecting anyone with a job or fellowship.
Underplaced searches from conventional public proof to unconventional indicators of intelligence.
Begin with small and midsized accounts. Their audiences are a free, noisy signal that an existing algorithm found something—but not proof, and not fame large enough to make discovery redundant.
Popularity is never a floor. A 60-follower reply account, obscure forum writer, or tiny game server may contain the strongest evidence. The crawler must search past self-promotion skill.
Apply directly or be nominated. Submit work samples and request the same assessment.
Follow a fictional grocery worker from the first public clue through assessment and a paid opportunity.
Paste a paragraph to see the simple language features that might prompt a closer human look.
Explore the technical design: how the search, evidence review, assessment, and placement pieces fit together.
See the complete data pipeline or a prototype that distinguishes evidence of technical judgment from popularity on GitHub.
Prototype status: Underplaced is a research project, not an operating hiring service. This site uses fictional candidates and does not crawl or score real people. Public evidence can prompt a closer look, but only a candidate’s voluntary participation and assessment can move the process forward.