AFD, Bibliotheques Sans Frontieres, Kajou and Pleias piloted Karibu, a French app built on a frugal 7B-parameter open model, with 30 teachers in Dakar/Thies. It gives level-matched exercises and feedback; local-language versions are planned; no outcomes published yet.
30
Teachers in initial pilot (Feb-Mar 2025)
Details
Maturity
Pilot
Promoter
Agence Française de Développement (AFD), with Bibliothèques Sans Frontières, Kajou and Pleias, in partnership with Senegal's Ministry of Education
Period
February-March 2025 initial pilot (ongoing feedback phase)
Karibu is an AI-powered language-learning application piloted in Senegal to help teachers strengthen their mastery of French, the country's official language of instruction, even though only a minority of teachers and pupils speak it as a first language. The pilot is funded by the Agence Française de Développement (AFD) and implemented with the NGO Bibliothèques Sans Frontières, the social enterprise Kajou and the research lab Pleias, in partnership with Senegal's Ministry of Education.
Activities
The application is built on a 'frugal' open-weight language model with seven billion parameters, chosen for Senegal's connectivity and cost constraints rather than a large proprietary model. It generates practice exercises matched to each user's level and stated interests, and gives detailed, individual feedback on users' written French. The initial pilot ran from February to March 2025 with thirty elementary-school teachers in Dakar and Thiès.
Results
As of this review, AFD had not yet published outcome data from the pilot; the partners describe it as an early test phase and say they are 'really looking forward to getting [teachers'] feedback' before deciding on wider rollout.
Conclusions
Plans for expansion include adding oral-conversation practice and versions of the app in Senegal's national languages — Wolof, Pulaar, Serer, Diola, Mandingo and Soninke — which would extend the tool beyond French-mastery support toward broader mother-tongue and multilingual education equity.
Implementation
Indicative cost
Low (< €50k) — Funded by AFD; specific budget figure not published. Model choice (7B-parameter open-weight) is deliberately low-cost/frugal.
Time to results
Short (< 1 year) — Initial pilot ran February-March 2025 with an ongoing feedback phase before any wider-rollout decision.
Staffing & skills
Agence Française de Développement (AFD) programme team, Bibliothèques Sans Frontières (implementation partner), Kajou (social enterprise), Pleias (research lab, model development), Senegal's Ministry of Education
Conditions for success
Choice of a lightweight, 'frugal' open-weight model suited to Senegal's connectivity and cost constraints
Gathering direct teacher feedback before deciding on wider rollout
Do you run this practice?
Claim it —
verified implementers get a public contact pathway and can propose corrections.
Data sources
Where this practice's information was retrieved from, and when.
OpenAI-funded IRC chatbot 'aprendIA' gives teachers in Bangladesh's climate-vulnerable coastal districts lesson plans and disaster-risk content via WhatsApp; built in …
Microsoft Research India and Sikshana Foundation built an AI copilot that drafts curriculum-aligned lesson plans in Kannada/English for Karnataka government-school …