UNICEF WCARO NLP Landscape

West and Central Africa Language Technology
Last updated: 2026-04-23
Work in progress. Some data may contain inaccuracies. Contribute on GitHub

AdwumaTech AI

Startup emerging

Organization Information

Startup
2023
medium
emerging
closed
Accra, Ghana
Commercial revenue from AI training data services for Fortune 500 clients. Enterprise contracts for RLHF, data annotation, and African language AI.
β€”
adwumatech-ai
β€”

Coverage

Ghana

Description

AdwumaTech AI provides enterprise-grade AI training data infrastructure powered by African engineering talent. Core offerings include RLHF for code generation, reasoning data development, data annotation, and African language AI solutions. They cover 6+ African languages with native-speaker teams for dataset creation, RLHF cultural alignment, and model evaluation/red-teaming. Also operates Nokore AI, a deepfake detection solution trained specifically on African identities for KYC/identity verification. Positions itself as building Ghana's first world-class AI workforce, with 100+ local jobs created.

πŸ¦„ UNICEF Relevance

Relevant for UNICEF Ghana's language tech needs through their African language AI services covering major Ghanaian languages (Akan, Ewe, Ga, Dagbani). Their native-speaker dataset creation and model evaluation capabilities could support quality assurance of NLP tools. However, their primarily commercial and closed approach limits direct integration β€” they are a data services company rather than an open-source contributor.

Key People

Projects

Notes

Primarily an AI training data company serving enterprise clients rather than an NLP research or product org. Their African language work focuses on dataset creation and model evaluation services rather than building open models. Only two datasets published on HuggingFace (both MIT-licensed speech datasets), no public models. Deepfake detection (Nokore AI) is a separate product line not directly relevant to NLP/language tech. Website claims 50+ African languages but only 6 are explicitly named.

Last updated: 2026-02-06