Microsoft has launched Project Gecko in Kenya. The initiative is led by Microsoft Research Africa in Nairobi. It brings together partners from Microsoft Research India, the Microsoft Research Accelerator in the United States, Digital Green, and local collaborators. The goal is to build AI systems that work for populations that are underrepresented online.

Project Gecko begins with agriculture. Kenya depends on agriculture for livelihoods and GDP. Millions of smallholder farmers need practical, local support. Many farmers rely on oral advice and visual demonstrations. Most global AI systems are trained on English data. They perform poorly in local languages and contexts. Project Gecko aims to change that.

The core technology is MMCTAgent. MMCTAgent is a multimodal AI system. It analyses speech, images, and video to produce context rich answers. The model can break complex questions into smaller parts. It can verify its answers against local knowledge. It can point users to the exact moment in a video where a solution appears. This makes advice actionable and trustworthy.

A key feature is voice interaction. Research in Kenya showed farmers prefer voice over text. Local languages include Swahili, Kikuyu, Kalenjin, Dholuo, Maa, and Somali. These languages lacked robust speech recognition and text to speech tools. Project Gecko developed speech models from scratch. The team trained models on a large, crowd sourced dataset. The dataset contains about 3,000 hours of Kenyan speech. This work improves recognition and helps AI understand local accents and terminology.

Also Read ~ James Mwangi: The Kenyan Visionary Reshaping Banking, Opportunity, and Social Impact

Project Gecko also uses small language models. These compact models run efficiently on low cost devices. They offer good performance for specialised tasks. They help bridge the gap where internet and compute resources are limited. This approach is designed to support real world deployment across rural areas.

Digital Green’s FarmerChat is a core partner and content source. Digital Green has collected more than 10,000 agricultural videos in over 40 languages and dialects. Project Gecko links AI reasoning with this rich library. A farmer can ask a question in Kikuyu and get an answer in audio and video. The system will guide them to the exact clip that addresses their problem.

Field studies in Kenya show early success. The tests reported better accuracy and higher user trust compared to generic AI systems. Farmers received clearer, locally grounded answers. Extension workers found the tool useful for demonstrating techniques in the field. The model reduced the time it takes to find relevant information in long videos.

Microsoft has made MMCTAgent available on Azure AI Foundry Labs. The team also released open source code for parts of the system. The move aims to accelerate developer adoption and local innovation. A public leaderboard is planned to benchmark African language performance. This will encourage wider participation from researchers and startups in Kenya and beyond.

The project offers direct benefits for Kenya. First, it can increase agricultural productivity by providing timely, accurate advice. Second, it can improve inclusion by serving farmers who use local languages. Third, it can strengthen local value chains by making best practices widely available. Fourth, it can create opportunities for Kenyan tech firms and startups to build services on top of the platform.

There are also challenges and risks. Connectivity in rural Kenya remains uneven. Many farmers use low end phones and limited data plans. The models need careful optimisation to work offline or with low bandwidth. Data privacy is another concern. The project must protect farmers’ personal data and ensure informed consent for recordings. Bias is a risk as well. The models must reflect diverse local practices and not favour specific crops or methods.

Scaling the system requires strong partnerships on the ground. Microsoft must work with extension services, local NGOs and county governments. Training for extension workers will be essential. Farmers need user friendly interfaces and clear guidance on how to trust and use AI advice. Payment models will also matter. Services must be affordable or subsidised for low income users.

Policy makers in Kenya can support the rollout. Investment in rural connectivity remains a priority. Policies that encourage public private partnerships can accelerate deployment. Kenya’s innovation hubs and universities can contribute to local research and help validate tools. The private sector can build services that integrate MMCTAgent into weather advisories, market price alerts, and input supply chains.

Project Gecko points to a broader shift in how AI is designed and deployed. The focus moves from global models to locally grounded systems. Microsoft’s work in Kenya shows that local language datasets, multimodal reasoning, and practical partnerships matter. The approach can extend to health, education, and other sectors. For Kenya, this is an opportunity to lead in locally relevant AI.

In short, Microsoft’s Project Gecko addresses a clear gap. It brings AI tools into the language and cultural contexts of Kenyan farmers. The technology can help improve yields, reduce crop losses, and boost livelihoods. With careful implementation, strong partnerships, and policy support, the project could become a model for AI for the global majority.

Source ~ Business Today