Data Methodology

DRAFT: NOT LEGAL ADVICE. Requires review by a qualified lawyer (India + EU/US exposure) before publishing. Operator: Anvith Shetty, an individual (sole proprietor, unregistered) in Bengaluru, Karnataka, India. Placeholders: The AI Race (AI Race HQ), [postal address pending], hello@airacehq.com (placeholder), 15 October 2026.

What's included: organizations whose core product or research is AI/ML (models, infrastructure, applications, robotics, AI services), with a working website and an identifiable location.

Sources (examples): company websites and public job boards; Wikidata (CC0); public registries (SEC EDGAR, UK Companies House, India MCA); GitHub and Hugging Face; reputable news/RSS; public social posts; company self-reports via claimed profiles. We do not scrape sites whose terms forbid it.

Cross-matching: every fact is stored with its source URL and date. Each source has a weight (company-verified 1.0, registry 0.9, company site 0.8, Wikidata/major news 0.7, other 0.4). A fact is published only when agreeing sources reach a score of 1.2, roughly two independent sources. Funding amounts are never published on a single source. Conflicts are shown as ranges or held for review.

Entity resolution: companies are identified by their primary web domain, with fuzzy name + location matching and Wikidata IDs to merge duplicates.

Freshness: funding and headcount are rechecked every 30 days, other fields every 90 days, sponsored profiles every 14 days. Each field shows "last checked".

Map: pins show the city of the HQ (not exact premises) unless a company verifies an office. Bubble size = headcount band (or total funding where headcount unknown), never payment.

Categories: a fixed taxonomy (~60 categories) assigned by classifier + review; companies can propose changes.

AI use: we use AI models to extract and summarize information, constrained to sourced facts; descriptions are written fresh, not copied. Errors can be reported anywhere on the site.

Reuse: cite as "The AI Race (AI Race HQ), 15 October 2026, [URL]". Aggregate city counts are available as CSV under CC BY 4.0.