Programmes and Projects
ICAIN’s pilot projects and funding programmes focus on four key areas in which AI can deliver real-world impact. The projects range from AI tools for more accurate weather forecasts and early detection of crop diseases in Africa to AI training for students in underresourced regions.
ICAIN's Key Areas
Climate & Agriculture Programme
In sub-Saharan Africa, millions of people depend on agriculture for their livelihoods, yet they face challenges specific to the region. Current pilot projects therefore aim to support local farmers with a focus on improving weather predictions for Africa and early disease detection in crops.
Early Diagnosis of Crop Diseases
As much as 40 percent of global crop production is lost to pests and sub-Saharan Africa is the world’s most vulnerable region to the risks of agricultural pests and diseases. Now, external page Dr Godliver Owomugisha at Busitema University in Uganda and her co-workers have developed a method to detect plant diseases before visible symptoms appear. The technique uses a smartphone add-on spectrometer that can be 3D-printed, creating a portable device to support farmers and experts working in the field. The team has completed the proof of concept and is currently validating the method in the field with local implementation partners.
Improved Weather Predictions for Agriculture in Africa
Accurate weather forecasts improve productivity and decision-making in agriculture, which is especially relevant in Africa, where the livelihoods of millions depend on rain-fed farming. But AI-based forecasts are less accurate for Africa than for Europe, for example, as this pilot project led by external page Ciira wa Maina, Associate Professor at Dedan Kimathi University of Technology in Kenya, has already shown.
This indicates that Africa is underrepresented in current AI prediction models. To improve their performance, the team’s next step is to fine-tune a widely used model with data collected by more than 700 weather stations across the continent.
The projects are supported by Fürstentum Liechtenstein.
IGAIP: an AI Boost for International Organizations
AI is advancing rapidly, yet many UN agencies and other international organizations lack access to sovereign computing resources, applied AI research capabilities, and reusable digital components. The ICAIN International Geneva AI Innovation Programme, IGAIP, addresses this gap. It connects ICAIN's academic network and large-scale computing resources with UN agencies and other international organizations based in Geneva to co-create AI solutions for global humanitarian challenges.
Current Projects
AI-Supported Interviews and Job Matching
Team: ETH Zurich, Prof. Dominik Hangartner together with the external page UN Refugee Agency, UNHCR
AI for Real-Time Integrated Genomic Analysis
Team: ETH Zurich, Prof. Tanja Stadler together with external page WHO
AI-Powered Early Warning and Profiling for Displaced Populations
Team: ETH Zurich, Dr. Christina Humer, Dr. Rita Sevastjanova together with the external page International Organization for Migration, IOM
Empowering Training, Education and Learning with AI
Team: ETH Zurich, Dr. Gerd Kortemeyer, Prof. Mrinmaya Sachan; EPFL, external page Prof. Martin Jaggi together with the United Nations Institute for Training and Research, external page UNITAR and the United Nations International Computing Center, external page UNICC.
An Ethical LLM for Humanitarian Action
Large language models (LLMs) could greatly benefit humanitarian organisations such as the external page ICRC. But in most existing models, humanitarian contexts are underrepresented. And the sensitivity of humanitarian data additionally limits use of off-the-shelf LLMs. In this project, a collaboration between the ICRC and AI scientists investigates the development of LLMs specifically tailored to the ICRC’s protection work for people affected by armed conflicts. An open-source suite of LLMs called Legitron is being developed in external page Prof. Mary-Anne Hartley’s group at EPFL, which will specialise in humanitarian law and be fine-tuned for domain-specific tasks such as legal reasoning.
The IGAIP programme is supported by the Swiss Department of Foreign Affairs, FDFA, and runs under its patronage.
Access to an open, multilignual LLM
So far, the global AI ecosystem has been strongly biased towards English and a small number of other high-resource languages. Underrepresented languages—the daily reality of billions of people—have largely been excluded from frontier AI development. ICAIN aims to help correct this by advancing open-science multilingual models, inclusive datasets, and governance frameworks that place linguistic equity at the centre of AI development. A perfect example is external page Apertus, developed by the ICAIN founding institutions ETH Zurich, EPFL and CSCS as part of the external page Swiss AI Initiative. Trained on 15 trillion tokens spanning more than 1,000 languages, Apertus is the world’s first truly open and multilingual LLM. It’s entire development process, including its architecture, model weights, training data and methods, is openly accessible and fully documented.
A joint evaluation of LLM safety
Facilitated by ICAIN, the United Nations International Computing Centre, external page UNICC, and the external page Machine Learning and Optimization Laboratory of EPFL joined efforts to enable safe, reliable and responsible use of LLMs. The project evaluated the open, multilingual LLM Apertus system across multiple dimensions, including its resistance to harmful prompts, its ability to recognize when it should not answer, the accuracy of its responses when using external knowledge, and its susceptibility to biased or misleading interactions. In addition, the project developed a practical evaluation framework that the public sector, the UN system, and other international organizations can adapt to assess the safety and reliability of their AI systems.
AI Training & Capacity Building
Across Africa, there is a significant gap between AI innovators and the communities intended to benefit from these technologies. That's why these pilot projects aim to improve AI literacy and foster academic exchange.
In addition, we seek to enhance the visibility and accessibility of training opportunities in supercomputing and AI, including summer schools, masterclasses, exchange programs, and training on large-scale computing and AI infrastructures.
AI for all
While interest in AI is growing across Africa, many communities still lack the foundational knowledge, skills, and computing infrastructure needed to engage with it in a meaningful way. The external page Busitema University AI and Interdisciplinary Research Group is working to bridge this gap by engaging diverse stakeholders across the education spectrum, from secondary schools and tertiary institutions to universities and policymakers. Through targeted training and collaboration, this pilot project aims to empower individuals and institutions to harness AI for education and innovation, and ultimately build technologies that address African challenges and improve livelihoods.
AI Driving Licence
Imagine learning about AI technologies through a game. The AI Driving License is a gamified, open-source education initiative that aims to empower citizens—especially young people—to build essential AI literacy for democratic participation. Through an interactive card game, players explore core AI concepts such as machine learning, bias, data ethics, and automation while engaging with real-world scenarios and ethical dilemmas. The project includes pilot workshops across Switzerland and a train-the-trainer programme to support scalable, responsible AI education.
The projects are partly supported by Microsoft.