In October 2023, we launched our inaugural Llama Impact Grants program, looking for innovative proposals from around the world that used our open source AI models to address major challenges in education, the environment, and technological innovation. Today, we’re excited to announce the recipients of those grants, as well as the results of the Llama Impact Innovation Awards.

Wadhwani AI, Digital Green in partnership with One Acre Fund, and the Dana-Farber Cancer Institute will each receive $500,000 USD, and our runner-up Jacaranda Health will receive $300,000 USD to support their unique global projects.

• Wadhwani AI proposed an AI-enabled reading assessment to enhance English fluency and comprehension skills in public schools in Gujarat, India. This module will utilize Llama to generate level-appropriate paragraphs for students, allowing them to focus on and gain mastery of words that they find challenging. By fine-tuning Llama, the generated content will be culturally relevant, safe, and aligned with curriculum standards, addressing the diverse needs of students.
  
• Digital Green’s Farmer Chat project aims to develop a multilingual AI chatbot that provides customized, on-demand agricultural advisory services to small-scale farmers in Sub-Saharan Africa. Llama will be fine-tuned on a vast repository of agricultural data to adapt the model to the specific needs and contexts of different regions. Llama’s conversational capabilities will then be used to provide comprehensive and practical responses in a variety of languages including Hindi, Swahili, and Kikuyu.
  
• The Dana-Farber Cancer Institute (DFCI), based in the US, proposed a new open source AI platform to computationally match patients with cancer to clinical trials. They plan to use Llama 3 to summarize and comprehend unstructured clinical notes and unstructured clinical trial eligibility criteria, enabling rapid identification of appropriate trial options for individual patients. The new AI models will be integrated into DFCI’s existing open source MatchMiner platform and evaluated with clinical partners.
  
• Runner-up Jacaranda Health aims to expand its digital health tool called PROMPTS that uses SMS behavioral nudges and an AI-enabled clinical helpdesk to empower new and expecting mothers across Kenya, Ghana, and Eswatini to seek and connect with the right care at the right time. The ultimate goal is to make PROMPTS the go-to resource for personalized maternal and newborn health support across Sub-Saharan Africa.

We’re inspired by these and the many other diverse projects we’ve seen developers undertake around the world to positively impact their communities by building with Llama.

In addition to the Llama Impact Grants, we’re also excited to announce the winners of the Llama Impact Innovation Awards, which are intended to further identify and support organizations based in Africa, the Middle East, Turkey, Asia Pacific, and Latin America that are already using Llama to address local issues related to education, public services, economic development, and more. Award winners will receive up to $35,000 USD, matched with in-kind ad credits, to support their work with Llama.

• Mohamed bin Zayed University of Artificial Intelligence’s BiMediX2 is a bilingual biomedical expert system that provides virtual medical advice and diagnostic support in both English and Arabic. Built with Llama 3.1, it can understand and respond to medical queries, including those related to medical images, and enables verbal user interaction by integrating pre-trained speech-to-text and text-to-speech models like Meta's Seamless-M4T to their system.
  
• HelpMum has developed an innovative chatbot service that delivers accurate and timely information on vaccination and immunization. Leveraging Llama 3’s advanced language processing capabilities, HelpMum created Vax-Llamamodel—an AI-driven, open-source model designed to provide reliable vaccination information. This open source model can be integrated into HelpMum’s chatbot service and adopted/used by healthcare organizations worldwide.
  
• Tanzania AI Community’s chatbot Twiga provides teachers with a virtual assistant for generating exams, lesson notes, and course materials tailored to the Tanzanian curriculum via WhatsApp. Twiga uses Llama 3 to generate contextualized educational content in order to address the shortage of quality resources and high workload of teachers in Tanzania. It uses Llama 3 70B in its RAG pipeline to generate high-quality educational exercises and Llama 3 7B for smaller tasks like efficient query pre-processing and data extraction.
  
• Flow Informatics’ phosoAI project aims to address food insecurity in Malawi by enhancing a WhatsApp-based chatbot that provides real-time information on food prices, availability, and agricultural practices to the public, particularly in rural and urban areas. The chatbot uses Llama 3 to provide real-time information to users in Chichewa. It also allows for voice commands to improve accessibility in low-literacy areas.
  
• Factly Media and Research’sSACH (Search Application for Claims & Hoaxes) is an AI chatbot designed to combat misinformation and enhance online trust and safety. SACH provides instant fact-checking capabilities and contextual insights for web content, including images and text. SACH’s multilingual capabilities ensure users can access fact-checking resources in their preferred language. Llama 3’s conversational abilities let users engage in natural language dialogues, ask follow-up questions, and explore topics in-depth.
  
• AIM Intelligence’s Llama-Suho project addresses AI safety and security challenges in the Korean context, focusing on culturally sensitive topics, nuanced societal issues, and hate speech. It leverages and fine-tunes Llama with Korean-specific data to enhance AI safety measures for the Korean language and cultural contexts. The project aims to contribute to the global discourse on ethical AI deployment and set new standards for culturally aware AI safety measures.
  
• Netsafeused Dagster as a foundation to design a robust data pipeline to help source, redact, and aggregate data in order to assist with harm remediation. NetSafe uses Llama 3 to analyze reports of online harm, redact sensitive information from reports in order to protect user privacy, and automate tasks like data aggregation, summarization, and categorization, freeing up staff to focus on high-priority cases.
  
• AiSee is an affordable wearable assistive device that helps people with visual impairment “see” the objects around them. The Llama 405B model serves as the backbone of the system, providing accurate and comprehensive information, while the 8B model, optimized for mobile devices, handles user interactions and generates intermediate responses.
  
• Hiperderecho’s aLex (Augmented Legal Experience) uses Llama to democratize access to and understanding of the Peruvian legal system. The team automated 60,000 PDFs of judgments available on the Constitutional Court portal from 1996 to 2024 in order to render the text searchable. Llama was used to classify, index, and make the text searchable for queries. Llama also creates summaries in both formal, legal language and informal synopses to help both those in legal professions as well as the public better understand the legal system.
  
• Instituto de Inteligência Artificial na Saúde’s NoHarm Summary Discharge project automates and enhances patient discharge summaries, improving accuracy, completeness, and efficiency. By streamlining the discharge process, the project reduces administrative burdens on healthcare providers, minimizes errors, and ensures comprehensive post-hospital care instructions for patients. This contributes to better patient outcomes, reduced healthcare costs, and improved economic efficiency in the healthcare sector.
  
• Civica Digital’s TrueDocs AI assistant provides real-time guidance and support for navigating government procedures, simplifying complex legal and technical language to make public services more accessible and user-friendly. The AI agent assistant simplifies the interaction with government services by providing real-time, multilingual support, breaking down complex procedures into understandable steps. This not only makes services more accessible but also increases the overall user satisfaction and trust in the system.

We’re committed to open source and believe that it’s the right path forward for AI. That’s why we’re working to expand access and opportunity to organizations across the globe.

In less than a year since we launched the first call for proposals for our inaugural Llama Impact Grants, we’ve expanded that original grant program into a portfolio of grants, awards, hackathons, and community building programs to showcase how Llama can accelerate efforts to promote social and economic progress. And we can’t wait to see what the community builds next.

Apply to the Llama 3.1 Impact Grants by November 22, 2024.