By Eddah Waithaka
A new initiative is building artificial intelligence that understands local languages and cultures, aiming to close the equity gap for the “global majority.”
Generative AI powers productivity tools across the globe, but these systems often fail communities underrepresented online. Because most AI training data originates from a narrow slice of the internet, the technology frequently performs poorly in many languages and ignores the social and cultural realities of billions of people.
Now, a Microsoft Research-led initiative called Project Gecko is tackling this disparity head-on. The project is assembling experts from Microsoft Research Africa in Nairobi, Microsoft Research India, and the Microsoft Research Accelerator in the United States, alongside partners like Digital Green, agri-tech leaders, and academics. Together, they are building cost-effective, tailorable AI systems that speak local languages, incorporate culturally relevant knowledge, and engage users through text, speech, and video.
“Building AI systems from the ground up shaped by the knowledge, languages, and modalities of the global majority yields more innovative, useful solutions for a great number of people,” said Ashley Llorens, Corporate Vice President and Managing Director of the Microsoft Research Accelerator.
Why Agriculture is the First Frontier
While Project Gecko plans to expand into healthcare, education, and retail, the team chose agriculture as its launchpad. In nations like Kenya and India, the sector employs millions of smallholder farmers who often work in linguistically diverse environments, switching between languages and relying on oral and visual instruction.
Existing agricultural apps frequently deliver incomplete or inaccurate answers because their underlying models train primarily on English data.
“Agriculture has very specific terms, which may change from language to language, and even district to district. All those domain-specific nuances need to be understood,” explained Tanuja Ganu, Director of Research Engineering at Microsoft India.
Project Gecko builds directly on Digital Green’s Farmer
Chat, a speech-first assistant that extension workers already use to support millions of farmers. Digital Green has amassed over 10,000 agricultural videos in more than 40 languages, but until now, linguistic and technical barriers locked away that knowledge.
“Unlocking this knowledge will support even more farmers to get real-time responses in their local language and preferred modality,” said Rikin Gandhi, CEO of Digital Green.
A Smarter, More Critical AI Agent
Powering this effort is a new multimodal AI system called the MultiModal Critical Thinking Agent (MMCTAgent). This system analyzes speech, images, and video to generate context-rich, locally grounded answers.
It breaks complex questions into smaller parts, verifies its own reasoning, and grounds its responses in real-world practices captured in community-generated videos.
Field studies in Kenya and India show the system significantly boosts accuracy, usability, and user trust compared to generic AI. A farmer in Nyeri County, Kenya, can now ask a question verbally in Kikuyu and receive a multi-modal answer, jumping directly to the exact timestamp in a video where the solution appears.
Lakshmi Devi, a farmer from Bihar, India, confirmed the impact: “Before, we would ask neighbors or dealers for advice and weren’t sure it was right.
With FarmerChat, we ask our questions, follow the instructions, and see better results.”Engineering a Voice for Local LanguagesMicrosoft’s human-centered research revealed a strong farmer preference for voice interactions.
However, many local languages lack the basic tools for automatic speech recognition and text-to-speech.The Project Gecko team is therefore building these tools from scratch, training them on local datasets.
Also Read : https://africawatchnews.co.ke/nba-returns-to-supersport-debuts-on-showmax-in-landmark-deal-for-african-fans/
To ensure the AI runs on low-cost devices common in rural areas, the team uses small language models (SLMs) compact, efficient models that often outperform larger ones in specialized, local tasks.
The team has already expanded support to Swahili, Kikuyu, Kalenjin, and Somali using a dataset of 3,000 hours of crowd-sourced Kenyan speech and is creating a public leaderboard to benchmark African language performance.
A Blueprint for Global AI
Project Gecko underscores a fundamental shift in how Microsoft and its partners approach AI localization. The team will soon release a multilingual playbook, offering practical guidance for developers creating tools for the global majority.
“Our goal is to ensure that the next generation of AI is not only powerful, but also globally inclusive, culturally relevant, and shaped by the communities it aims to serve,” Ganu emphasized.
By starting in the fields of Kenya and India, Project Gecko is planting the seeds for a more equitable AI future, one where technology truly speaks everyone’s language.
