BUSINESS CASES · PRODUCTION DEPLOYMENTS
Six pipelines where Mondonomo's name infrastructure replaces brittle rules, fragile lookup tables, and hallucinating LLMs. Each case shows the model composition, the metric that moves, and which industries pay for it.
Six concrete revenue paths against the $12.9M / org / yr cost of bad name data (Gartner). 99% of language vocabulary is names — and most of it sits outside general LLMs.
Investor inquiry →Pilot the API at scale — KYC vendors, CDPs, identity platforms, LLM tooling. Co-built integrations, dedicated rate limits, joint go-to-market on cultural-data wins.
Request a pilot →One REST API replaces a stack of regex, lookup tables, and LLM prompts. Same key works across all five Nelma surfaces. Sub-50ms typical latency.
Try the API →Every pipeline cites the underlying paper and evaluation methodology. Reproducible eval scripts; Hugging Face weights for the published components.
Read the papers →A sanctions list contains يوسف بن أحمد. A wire transfer comes in for Yousef Ben Ahmed. A second comes in for Yusef Bin Achmed. Are these the same person? Naïve string matching says no. A general-purpose LLM hallucinates a confident wrong answer. MondoGraph + MondoPhon resolves them as a single IPA cluster with auditable evidence — the regulator can see every step. "$12.9M average annual cost of poor data quality per organization." — Gartner, cited in the SmartBase pitch deck.
The traditional sign-up form asks for First / Middle / Last Name. That works for ~30% of the world. For Maria Teresa García Ramírez de Arroyo, every traditional form either truncates her, mangles her, or destroys her maternal lineage. Mondonomo's Smart Forms product takes one free-text field, runs Classify → Parse → Gender → Transliterate, and lands a structured, culturally-correct record with the right salutation for the marketing email that follows. "With one field, SmartForms analyses all information for you." — 2024 Mondonomo pitch deck.
Sales reps type names the way they hear them. Yousef, Yusuf, Yousef Ahmed, Joseph, يوسف — five rows in your CRM, possibly one person. MondoPhon's IPA2vec embeddings cluster phonetic variants across writing systems; the parser splits each row to compare given vs. surname independently. Run once on existing data, then on every new lead.
"Dear MARIA TERESA GARCIA RAMIREZ DE ARROYO" is the marketing email that gets unsubscribed. "Dear Señora García," is the one that gets opened. The same applies to Khun Chomphunut, Mǎ Yún Xiānshēng, and سامي بن أحمد. PNEUMA-DD + MondoPhon produce the right form for every customer's locale — formal and casual variants, in their native script — from a single name string. "To: Maria Teresa García Ramírez de Arroyo · Dear Señora García / Hi Maria." — Mondonomo Smart Forms pitch.
A patient registers at hospital A as Mohammed Al-Saud; at hospital B as محمد آل سعود; at the social-security ministry as Mohamed Alsaud. Three records, one citizen, three care plans that don't see each other. MondoPhon's transliteration ensemble normalizes across scripts; PNEUMA-DD parses and confidence-scores each match. Used at population scale, this is the difference between a fragmented and a unified national identity system.
Frontier LLMs don't know what 99% of your customers' names mean. They guess. They hallucinate genders,
mispronounce names in voice agents, confuse two unrelated people. Mondonomo's API exposes a single
/enrich endpoint that any LLM can call as a tool:
pass a name string, get back structured facts (type, gender, country, transliterations, pronunciation,
known bearers) to inject into the prompt context. Designed for voice agents, customer-support bots,
and any application where getting a name right matters.
"99% of an average language vocabulary consists of proper names. Most are not within the grasp of large language models." — Mondonomo, 2024.
ADJACENT · CULTURAL & RESEARCH PROJECTS
Mondonomo's consumer surfaces aren't just marketing — they're how MondoGraph grows. Every search, every rating, every contribution adds rows.
The flagship consumer site. Type your name, see prevalence maps, variants, notable bearers, and articles. Source of community contributions.
The Thai-focused prototype. Published with Prof. Attapol Rutherford (Chulalongkorn). Includes Maae Cham, the bilingual AI genealogical assistant.
Find names that sound like yours in other languages. The most direct showcase of MondoPhon + IPA2vec. Pure delight; rich data signal.
TALK TO US
We work with banks, e-commerce platforms, public-sector data teams, LLM companies, and academic groups. Tell us your name problem; we'll show you the simplest path to fix it.