Researcher Designs Privacy-Conscious AI for Smart Spaces

Most smart building systems today rely on closely monitoring people in order to understand how spaces are used. This work is grounded in a different philosophy: intelligent and responsive buildings can be designed in ways that strongly respect human privacy.
By using rich, privacy-preserving data, graph-based deep learning models are being developed to detect occupancy, estimate the number of people in a space, and infer basic activities such as sitting, walking, presenting, or group discussions. These insights are then linked to indoor air quality and energy consumption, enabling smarter and more efficient control of ventilation and HVAC systems.
The engineer explained that this research represents a new direction for smart buildings spaces that adapt to people in real time, improve health and comfort, and significantly reduce energy waste.
Emphasising the relevance of privacy-preserving activity sensing for Nigeria and across Africa, he noted that such systems are better aligned with local values and realities. Many communities are understandably wary of heavy surveillance, and solutions that rely on non-camera sensors can build trust while still improving comfort, safety, and efficiency.
Energy costs are often high and power supply can be unstable. As a result, schools, hospitals, and offices stand to benefit greatly from buildings that automatically adjust ventilation and cooling based on real activity rather than fixed schedules, reducing energy use while maintaining healthy indoor environments.
He added that just as African countries have leapfrogged to mobile banking and distributed solar power, they can also leapfrog to privacy-respecting smart buildings instead of adopting camera-heavy systems developed elsewhere. The methods being developed graph-based models, multimodal sensing, and privacy-aware algorithms can be adapted to low-cost sensors and deployed in classrooms, clinics, and offices across the continent.