Meta, in partnership with Data Science Africa, has unveiled the winners of the 2025 Llama Impact Grant for startups and researchers, awarding $20,000 in funding to each recipient alongside mentorship, policy engagement, and ecosystem support to scale their solutions.
The announcement was made during the Unstoppable Africa 2025 event at the United Nations General Assembly in New York.
The initiative, launched in March 2025 as part of Meta’s global Llama Impact Grants programme, is designed to enable innovators across Sub-Saharan Africa to deploy the company’s open-source large language model, Llama, in solving pressing social and economic challenges. With a strong focus on healthcare, agriculture, education, and digital accessibility, the programme builds on Meta’s broader commitment to strengthening Africa’s AI and innovation landscape.
Sherry Dzinerova, director of AMET Public Policy, Programs, Campaigns and Product at Meta, described the quality of applications as evidence of Africa’s growing AI ecosystem. “We received an incredible number of applications this year, reflecting the vibrant and growing AI ecosystem across Africa. These projects exemplify the spirit of innovation and impact that the Llama Impact Grant for Startups and Researchers stands for. We are excited to support their journeys and look forward to seeing the positive change they will bring to their communities and beyond,” she said.
The winners of the 2025 Llama Impact Grant include Vambo AI from South Africa, led by co-founder and CEO Chido Dzinotyiwei, which is developing multilingual AI infrastructure for Africa using proprietary and open-source models to power translation, transcription, content generation, and search across more than 60 African languages, aiming to treat language as digital infrastructure to drive inclusion and accelerate innovation.
PropelMapper, also from South Africa and founded by Reghardt Adriaan Pretorius and Mark Donne, provides agriculture advisors with AI-powered tools, including tailored farmer podcasts, automated reporting from debriefs, and satellite-driven intervention alerts using Llama models, to improve productivity and food security.
Radease from Nigeria, co-founded by Taiwo Oyewole, simplifies access to safe medicines by equipping Patent and Proprietary Medicine Vendors with WhatsApp-based AI tools, enhancing the availability of verified health information and essential resources, particularly in underserved communities.
TeenApp from Uganda, developed at Makerere University’s AI and Data Science Center under lecturer and researcher Rahman Sanya, delivers accurate, youth-friendly sexual and reproductive health education guided by Responsible AI principles to provide accessible, trusted digital health services for young people.
Easy Read Africa from Rwanda, led by Isaac Manzi, AI and Data Innovation Advisor at GIZ and Community Manager at MbazaNLP, converts complex documents into simplified text, visual explanations, and natural voice narration, expanding access to information for individuals with cognitive and learning challenges.
