SA study tool wins Meta Llama Impact Grant

Ten international recipients of the second Llama Impact Grant have been named, and they include a South African platform that provides students with educational resources through messaging apps.

The local winner, FoondaMate, was inspired by co-founder Dacod Magagula’s experience growing up in a rural township in Mpumalanga, where limited textbooks and overcrowded classrooms made learning challenging.

The multilingual study tool, based in Sub-Saharan Africa, provides exam preparation and schoolwork support to four million students with limited resources. The platform operates through chat interfaces on WhatsApp and Facebook Messenger and is available in 10 languages across more than 30 countries, including Colombia, Mexico, Brazil, Indonesia, and regions in Sub-Saharan Africa.

The grants make up over $1.5-million USD in awards and are intended to spur innovation among companies, startups and universities using Llama to create economic and social value.

Llama’s open-source framework is free to use, providing a resource for developing advanced technology and fostering economic growth. This accessibility enables organisations to create tools and launch businesses, supporting the development of cost-effective, high-powered solutions for emerging technologies.

The second Llama Impact Grants highlight companies, startups, and universities leveraging Llama to drive impactful change, from providing AI insights for farmers to enhancing fraud detection in the digital economy. These organisations are developing technology aimed at shaping the future.

“Open models such as Llama have shown that they can produce high-quality feedback, support multilingual interactions and enable novel learning activities,” says one of the winners, the University of Auckland. “For us, this means we are not forced to trade-off between performance and openness – we can achieve both.”

The first Llama Impact Grant, announced in September 2024, demonstrated the potential of open-source AI to impact sectors such as healthcare, education, and agriculture. This year, interest has increased, with over 1,300 applications received from more than 90 countries.

The Llama Impact Grant recipients include:

Doses AI (UK) leverages Llama in building an autonomous pharmacy system that transforms traditional medication dispensing and pharmacy operations, while maintaining pharmacist final check for patient safety. With Llama’s advanced vision and text processing system, the technology can process prescriptions, automate stock ordering, guide robotic retrieval systems to match prescriptions and detect potential errors in real time. The grant will accelerate the development of its LLM-powered robotic dispensing system.  

The University of Padova’s (Italy) TaccLab research group is using Llama to transform antibiotic discovery and generate new antibiotic molecules with the help of large datasets of chemical and biological sequences. The grant will help make AI-assisted drug discovery and material innovation more efficient, enabling the validation of new chemical compounds while speeding up research timelines and reducing costs.

Counterfake (Turkey) provides an online brand protection solution. The organisation uses Llama models in their pipelines to catch counterfeiters, helping brands protect their reputation from fraudulent online products, and consumers protect themselves from falling victim to ecommerce scams. The grant will help develop an API solution for counterfeit detection.

EERS (US) is a Llama-powered chatbot developed by entertwine that helps people navigate public services. For the chatbot’s initial phase, entertwine has collaborated with the Medi Community Resource Center – a nonprofit that works to connect communities to healthcare and social services – to create a solution that can be easily integrated with government databases. The grant will help scale and pursue these integrations. 

Solo Tech (US) uses Llama to offer offline, multilingual AI support for underserved rural communities with limited internet access. This includes insights for farmers, medical assistance for doctors and learning solutions in schools. The grant will help to equip 50 rural centers with AI tools while empowering local professionals and advancing the development of physical AI solutions.

Nova Escola (Brazil) is a non-profit organisation that equips public school educators across Brazil with digital tools and resources. Its AlfaTutor tool uses Llama to help teachers craft personalised lesson plans suited to the unique needs of each student and is accessible even in regions with limited internet connectivity. The grant will support the development and large-scale implementation of the solution.

BluEye (Mexico) is building, in partnership with LEX University, a mobile app to improve hurricane preparedness and response. Using real-time data from official APIs and Llama, BluEye provides personalised information, including early warnings, interactive maps and educational resources, that enable users to make informed decisions before, during and after a weather event, even in areas with limited connectivity. The grant will help improve offerings and expand geographic reach.

The University of Auckland(New Zealand) uses Llama to create accessible, multilingual tools that aid novice programmers and drive digital literacy. Its Explain in Plain Language (EiPL) and Prompt Problems project guides students as they develop foundational coding skills. Driven by AI code generation, students receive immediate, transparent and actionable feedback as they develop skills in code understanding and prompt engineering. With the grant, the University will explore novel feedback mechanisms, support additional programming languages and improve multilingual support, with an emphasis on Indic languages.

Nayana (India) has been developed by Cognitive Lab, an India-based research lab with an open-source-first approach, to give access to advanced AI capabilities to more than three billion people in underserved language communities globally. This project is a Multimodal Language Model that integrates Llama to automate complex document and image processing across multiple languages. The grant will support CognitiveLab in expanding Nayana’s language coverage and advancing its overall capabilities to serve low-resource regions more effectively.