I built a single power scale for people, companies and institutions (what worked, what broke)

Описание и применение

Создана платформа ApexHierarchy, которая ранжирует людей, компании и организации по единой шкале из восьми уровней и шести измерений, используя LLM для генерации текста и формулу для назначения уровня.

Задача

No unified metric to compare diverse entities such as city mayors, CEOs, ministries, and tech companies on a single scale.

Результат

777 profiles, 3,127 connections, 1,648 sources across 40+ cities, with a functional search, ranking, and connection mapping system.

Рабочий процесс

• 1. User searches for a person, company, institution or country; if absent, the system researches public sources in ~25 seconds.
• 2. An LLM writes the profile text and extracts figures.
• 3. A deterministic formula assigns the tier based on extracted data and reach caps.
• 4. The profile displays sources, corroboration flags, and a data‑confidence score.
• 5. Users can report mistakes, claim profiles, or submit missing ones.
• 6. Paid plans unlock additional features such as PDF export and watchlists.

Доказательства

Search any person, company, institution or country. If it isn't in the index yet, it's researched from public sources on demand (about 25 seconds).

Источник

Public sources → an LLM writes the text and pulls figures → a formula (not the LLM) assigns the tier.

Источник

777 profiles, 3,127 connections, 1,648 sources, 40+ cities.

Источник