Ethiopian Artificial Intelligence · Addis Ababa

The AI lab
of Ethiopia.

We build artificial intelligence for Ethiopian languages — speech, media and models designed in Addis Ababa, deployable anywhere in the country. In-house trained. Production proven.

$ voice.et transcribe --lang am call.wav
✓ 97.4% accuracy · Amharic transcribed in real time
 
$ huggingface download Lab-et/amharic-corpus
✓ 146,359 clips · 408.5 hours · open, CC BY 4.0
 
$ huggingface download Lab-et/azemari
✓ azemari · amharic · afaan oromoo · tigrinya · open TTS model
 
$ who we are
an ethiopian ai lab
408.5hopen amharic speech
5languages in production
193ktraining utterances
2live products
01

Two products. One lab.

02

What we do for organizations.

From a single assistant to a sovereign AI platform — we design, build, train and run AI for Ethiopian organizations, in Ethiopian languages.

01

Language & voice AI

Speech and language AI that understands Ethiopia.

  • Speech-to-text in Amharic, Afaan Oromoo, Tigrinya and more
  • Natural text-to-speech and custom brand voices
  • Translation & localization
  • Call-center voice bots and IVR
  • Call transcription & analytics
  • Document reading (OCR) for Ge’ez script, IDs and forms
02

Custom models & fine-tuning

AI trained on your data, for your work.

  • Fine-tuning language and speech models on your data
  • Private AI assistants that answer from your documents
  • Classification & data-extraction models
  • Domain vocabularies for banking, health, legal and more
  • Deployed on your servers or hosted in Ethiopia
03

Workflow automation & AI agents

Hand repetitive work to AI — keep people for what matters.

  • Customer support on Telegram, web and phone
  • Document, form and report processing
  • Meeting & call notes, summaries and action items
  • Complaint routing, email drafting and HR screening
  • AI agents connected to your CRM, ERP, email and Telegram
04

Data & AI infrastructure

The foundation your AI runs on.

  • Ethiopian-language data collection & labeling
  • AI API platform hosted locally, billed in Birr
  • On-premise & sovereign AI deployments
  • Data analytics & dashboards
  • Computer vision — ID checks, inspection, satellite & CCTV analytics
05

Advisory & training

Helping leaders and teams use AI well.

  • AI strategy & readiness assessments
  • AI policy, data protection & responsible-AI advice
  • Staff AI training in Amharic and English
  • Technical training for engineering teams
06

Research & partnerships

Moving Ethiopian AI forward, in the open.

  • Open models, datasets and benchmarks
  • Joint research with universities & institutes
  • Ethiopian-language AI for global companies
  • Hackathons, internships and events
03

Built for every sector.

Banking & fintech

Fraud detection, credit scoring, KYC document checks and Amharic customer assistants.

Telecom

Customer-care voice AI, call analytics and churn prediction.

Government

Citizen-service assistants, document digitization, archive search and translation of laws.

Health

Medical transcription, health-information assistants and record digitization.

Agriculture

Voice advisory for farmers, crop-disease detection from photos, weather and market-price alerts.

Education

Amharic AI tutors, exam preparation, automated grading and local-language content.

Media

Transcription, subtitles, dubbing and news summarization.

Legal

Contract review and legal research in Amharic.

Don’t see your sector? We build to fit.Tell us what you need
04

Our data is open.

open releases

Our Amharic speech corpus — open to everyone.

Generated entirely by our own OmniVoice TTS, deduplicated, transcribed and released under CC BY 4.0. Use it to build Ethiopian AI — and tell us what you make.

146,359 clips408.5 hours16 kHz wav62 speakersCC BY 4.0
05

Why lab.et.

Data stays in Ethiopia

On-premise and sovereign deployments aligned with the 2024 Data Protection Proclamation — built for banks, telcos and government.

Own models, own datasets

Models we train and adapt in-house, on datasets we collected ourselves. No rented intelligence, no foreign data flows.

Production-proven

Five languages live in production, 193k utterances in training, real products serving real users — not slideware.

Questions.

What is lab.et?

lab.et is the AI Lab of Ethiopia — a research and engineering company in Addis Ababa that designs, builds, trains and runs AI for Ethiopian organizations, in Amharic, Afaan Oromoo, Tigrinya and more.

What products does lab.et build?

lab.et builds and operates yakal.et — the Ethiopian AI console for chat, images, video and voice — and voice.et, its speech-recognition and natural-voice platform for Ethiopian languages, trained on its own corpora.

Does lab.et offer custom AI development?

Yes. lab.et fine-tunes language and speech models on your data, builds private assistants that answer from your documents, and deploys on your servers or hosted in Ethiopia — aligned with the 2024 Data Protection Proclamation.

What languages does lab.et support?

Five Ethiopian languages are live in production, including Amharic, Afaan Oromoo and Tigrinya — across speech recognition, text-to-speech, translation and Ge’ez-script OCR.

Does lab.et publish open data?

Yes. lab.et released an open Amharic speech corpus of 193,000 utterances — generated by its own OmniVoice TTS, deduplicated and transcribed — under CC BY 4.0.

How do I contact lab.et?

Email admin@lab.et or use the contact form at https://lab.et/#contact.

06 · start

Build with the lab that speaks your language.

Custom models, workflow automation, speech AI for your contact center, or just a question — send us a message, it lands directly in the lab's inbox.