Entity resolution as an API

Know when two names are the same entity.

One API call scores whether two names, in any script or spelling, refer to the same person or organization. Built for screening, investigation, and compliance at scale.

Get a sandbox key Read the docs
Live match
0.00
Demo runs a reduced model in your browser. The API adds learned embeddings, alias graphs, and transliteration coverage for 40+ scripts.
Why now
A deepfake attack every five minutes.

That was the detected rate across identity verification in 2024. Generative tools made convincing impostors cheap, and every fake arrives wearing a plausible name.

3,000%rise in deepfake fraud attempts in a single year.Onfido Identity Fraud Report
$40Bprojected US fraud losses driven by generative tools by 2027, up from $12.3B in 2023.Deloitte Center for Financial Services
42%of detected fraud attempts now involve AI-assisted identity manipulation.Signicat, 2024
The shift

The name is the last stable signal. Screening it still runs on rules written before Unicode.

Faces and documents can now be generated. Names still have to survive contact with watchlists, registries, and court records across 40+ writing systems. That check is broken.

Before Onoma

Rule engines guess.

  • Fuzzy-match rules, transliteration tables, and alias lists maintained by hand.
  • 95%+ of screening alerts are false positives an analyst still clears manually.
  • One spelling shift or script change and the real target walks through.
  • Every vendor scores differently. No threshold means anything.
After Onoma

One call resolves.

  • One endpoint replaces the matching stack: any script, any spelling, any alias.
  • Calibrated scores. 0.91 means the same thing on every pair, so thresholds automate.
  • Evidence with every score: script, token, and alias-level contributions auditors can read.
  • False positives drop without loosening the net.
How it works

One endpoint. A score you can act on.

  • Cross-script matching. Latin, Cyrillic, Arabic, and transliterated forms resolve to the same entity.
  • Calibrated scores. 0.91 means the same thing on every pair, so you can set thresholds and automate.
  • Explainable output. Every score ships with the evidence: script, token, and alias-level contributions.
  • Batch and stream. Screen one pair or ten million. Same endpoint, same latency budget.
$ curl https://onoma.kineticpartners.org/api/v1/match \
  -H "Authorization: Bearer sk_live_..." \
  -d '{
    "a": "Mohammed al-Husseini",
    "b": "Мухаммад Хусейни"
  }'

{
  "score": 0.94,
  "verdict": "match",
  "evidence": {
    "script_pair": "latin/cyrillic",
    "translit": 0.97,
    "token_align": 0.92
  }
}
Who it's for

Anyone who has to answer "is this the same person?"

Sanctions and KYC screening

Cut false positives without loosening thresholds. Score every hit against watchlists in any script, with evidence your auditors can read.

Investigators and OSINT

Collapse aliases, transliterations, and shell-entity name variants into single subjects across sources.

Fraud and risk platforms

Catch repeat offenders re-registering under name variants your string matching misses.

Data teams

Deduplicate entity tables across languages without building a matching pipeline from scratch.

Pricing

Priced per lookup. No platform fee.

Sandbox

Free
  • 1,000 lookups / month
  • Full API, test keys
  • Community support
Get a key

Production

$0.002 / lookup
  • Volume discounts at scale
  • Batch endpoint
  • 99.9% SLA
  • Evidence output
Request access

Enterprise

Custom
  • Private deployment
  • Custom alias corpora
  • Dedicated support
Talk to us
Access

The matching layer for the world's names.

Access is open to screening, investigation, and compliance teams.

Request access