Project Pegasus
Physical AI for aerial and legged autonomy
Project Pegasus is my current deep focus: bringing intelligence out of the cloud and into robots that move through real environments. The core idea of Physical AI is simple but hard - models that do not just reason over text or images, but close the loop on sensing, planning, and control under real-world physics, latency, and uncertainty.
On the drone side, Pegasus explores autonomous flight stacks for inspection, mapping, and rapid situational awareness. That means onboard computer vision, obstacle-aware navigation, mission planning, and edge inference that can keep flying when links drop - especially relevant for rural and infrastructure-scarce African settings.
On the quadruped side, the work centres on legged platforms that can traverse rough ground, climb slopes, and operate where wheels and fixed paths fail. I am building toward robust locomotion, multi-sensor perception (cameras, IMU, depth), and behaviours that let a four-legged robot patrol, inspect, or assist humans in dynamic outdoor environments.
Pegasus is systems engineering as much as AI: flight controllers and gait stacks, sensor fusion, simulation-to-real transfer, power and thermal constraints, and safety layers so these platforms can eventually be trusted in the field - agriculture, logistics, disaster response, security, and industrial inspection.
Aerial Autonomy
Drone perception, flight planning, and resilient autonomy for inspection and aerial intelligence missions.
Legged Robotics
Quadruped locomotion, terrain adaptation, and field behaviours for uneven and constrained environments.
Physical AI Stack
Onboard models, sensor fusion, and control loops that turn perception into reliable real-world action.
See my personal first-principles learning roadmap for Pegasus →