AI training data providers for Africa (2026): who does what.
A plain guide to companies, open datasets and communities that supply African AI training data, with each one’s public focus in a sentence or two and a link to the source. Listed A to Z within each group. Not a ranking.
Free 10-hour sample first. A person replies within one working day.
Companies that collect, label or license African AI data.
Each summary comes from the company’s own site or reputable press. Check their site for current services.
Awarri · Lagos, Nigeria
Builds voice AI and language-data platforms for Yorùbá, Hausa, Igbo and Pidgin, including Langeasy (crowdsourced language data) and RootLens (visual data), and offers labelling, annotation and RLHF services. The Federal Ministry of Communications, Innovation and Digital Economy announced N-ATLAS, an open multilingual language model, in September 2025 and named Awarri as the company powering it.
Source: Awarri website
Aya Data
Data acquisition and annotation across computer vision (including egocentric video), NLP, 3D and LiDAR, and medical imaging, plus RLHF and speech services (AyaSpeech).
Source: Aya Data website
CloudFactory · Founded in Nepal in 2010; offices in Nairobi and Kathmandu
Human-in-the-loop data labelling and AI lifecycle services with a managed workforce.
Source: CloudFactory: about
DataLens Africa · Lists 15 African delivery countries, including Nigeria
A human-in-the-loop data company from CipherSense AI offering data annotation, data collection (including speech), model evaluation and localisation, and says it covers more than 40 African languages. It launched an AI talent network in February 2026.
Source: DataLens Africa website
Digital Umuganda · Kigali, Rwanda
Voice and text data collection for African languages, and a data-collection partner on Google’s WAXAL speech project.
Source: Digital Umuganda website
GeoPoll · Denver, US, with offices in Nairobi, Johannesburg and Dar es Salaam
A research and data-collection firm that also licenses recorded, transcribed phone-conversation data from Africa, Asia and Latin America for LLM and voice AI, including Hausa and Swahili sets.
Source: GeoPoll: LLM voice data
Intron Health · Founded in Nigeria
Speech recognition and text-to-speech for African languages and accents (its Sahara platform), with a strong healthcare focus. Released AfriSpeech-200, an open accented-English dataset of clinical and general speech under CC BY-NC-SA 4.0.
Source: Intron website
Lelapa AI · South Africa
Builds language technology for African languages: Vulavula (speech-to-text with code-switching, and translation) for businesses such as call centres, and InkubaLM, a small language model for African languages.
Source: Lelapa AI website
Sama · San Francisco, US, with delivery centres in Kenya and Uganda
Human-verified data annotation, validation and model evaluation for image, video, LiDAR and text. Founded in 2008 as Samasource.
Source: Sama website
Sunbird AI · Kampala, Uganda
Speech and translation technology and datasets for Ugandan languages, including the SALT dataset.
Source: Sunbird AI website
TrustAm Data · Lagos, Nigeria
Us. Egocentric (head-mounted) video of skilled trades and Nigerian-language speech (Pidgin, Yorùbá, Igbo, Hausa, Nigerian English), collected to order from ID-checked, consenting recorders at workplaces registered in person, with every minute reviewed by a person. A Lagos pilot, scaling.
Source: TrustAm Data
Open African speech data worth starting with.
Open data is often the right first step. Read each licence: some forbid commercial use.
African Next Voices
An open speech-data project funded mainly by the Gates Foundation, led by Maseno University (Kenya), Data Science Nigeria and the University of Pretoria. Its Nigeria languages include Hausa, Igbo, Nigerian Pidgin and Yorùbá.
Mozilla Common Voice
Crowdsourced open speech data released under CC0, with volunteer-recorded sets for many languages, including Hausa, Igbo and Yorùbá.
Source: Mozilla Common Voice
NaijaVoices
An open speech dataset of Igbo, Hausa and Yorùbá from thousands of speakers, released under CC BY-NC-SA 4.0 (non-commercial).
WAXAL (Google Research and partners)
An open African-language speech resource (ASR and TTS) under CC-BY-4.0, collected with African partners including Makerere University, the University of Ghana and Digital Umuganda.
Where African AI data people meet.
Not vendors, but the best places to find collaborators, papers and datasets.
Data Science Nigeria · Lagos, Nigeria
An AI talent, research and community organisation founded in 2016, and the Nigeria lead of the African Next Voices speech project.
Source: Data Science Nigeria website
Masakhane
A grassroots research community for African-language NLP, behind projects such as MasakhaNER and Masakhane MT. Not a vendor, but the best place to meet people building African-language AI.
Source: Masakhane website
Where these facts come from.
- Awarri website
- Federal Ministry of Communications, Innovation and Digital Economy: N-ATLAS announcement
- Aya Data website
- Aya Data: about us
- CloudFactory: about
- DataLens Africa website
- BusinessDay, 18 Feb 2026: DataLens Africa launches AI talent network
- Digital Umuganda website
- GeoPoll: LLM voice data
- Intron website
- AfriSpeech-200 on Hugging Face
- Lelapa AI website
- Lelapa AI blog, Jun 2025: InkubaLM
- Sama website
- TechCrunch, Nov 2019
- Sunbird AI website
- TrustAm Data
- The Conversation (via Down To Earth), Oct 2025
- Mozilla Common Voice
- NaijaVoices dataset card, Hugging Face
- Google Research blog, Mar 2026: WAXAL
- WAXAL on Hugging Face
- Data Science Nigeria website
- Masakhane website
Buyers usually ask.
Is this list ranked?
No. Entries are A to Z within each group. Which provider is “best” depends on your data type, languages, countries and licence needs.
Why is TrustAm Data on its own list?
Because it is our page and we are one of the options. We describe ourselves with the same plain facts as everyone else, including what we do not do.
Which provider films egocentric video in Africa?
Aya Data lists egocentric video among its computer-vision services. TrustAm Data records head-mounted video of skilled trades in Nigeria. Check each provider’s site for current offerings.
Something here is wrong or out of date. How do I fix it?
Email [email protected] with the correction and a source. We check and update the page.
More from TrustAm Data.
Tell us what you need. We send a free sample first.
Email [email protected] with the data type, hours, languages or trades, labels and format. Pricing is quoted per approved hour after scoping.