Pioneering AI for Human Communication
The Inkulumo AI Research Lab advances deep learning to make human communication more accessible, inclusive, and intelligent across diverse languages and communities.
The Inkulumo AI Research Lab advances deep learning to make human communication more accessible, inclusive, and intelligent across diverse languages and communities.
Based in Johannesburg, the Inkulumo AI Research Lab is dedicated to solving complex human communication challenges. We bridge the gap between advanced technology and real-world accessibility through pioneering research and ethical, human-centric development.
Our work begins and ends with people. We are committed to a 'community-first' approach, developing responsible AI in direct partnership with the communities we serve to ensure fairness, transparency, and cultural respect in every model we build.
We pursue foundational research in areas critical to accessibility, including multimodal AI, low-resource language processing, and advanced computer vision, with a focus on African languages and sign languages.
We translate complex research into tangible, life-changing tools. Our work moves beyond theory to create practical applications, like our award-winning SASL translator, that empower individuals and foster a more inclusive society.
Our research agenda is driven by our mission to create accessible technologies. We focus on key AI domains that allow us to build bridges between different modes of human communication.
Developing models that understand and generate low-resource African languages and the unique grammatical structures of sign languages.
Creating robust systems for the real-time recognition of gestures, facial expressions, and complex sign languages from video.
Researching and developing realistic 3D avatars that can perform fluid, grammatically correct sign language animations from text input.
Building systems that can process and integrate multiple data streams—like video of signs, spoken audio, and text—for a comprehensive communication model.
Ensuring our models are transparent and trustworthy. We build systems whose decision-making processes can be understood by users and researchers.
Aligning our AI's objectives with human values and the well-being of the communities we serve, ensuring our technology is safe and beneficial.
Exploring privacy-preserving techniques to train powerful AI models on decentralized data, protecting the privacy of the communities we work with.
Dedicated to creating foundational language models and practical tools for the many under-represented and low-resource languages of Africa.
Interested in collaboration, research opportunities, or learning more about our work? We'd love to hear from you.