Research
An independent, early-stage exploration of AI systems that perceive, understand, and respond to the physical world.
What I am exploring
I am interested in the layer between AI models and the environments in which people actually live and work: cameras, sensors, edge devices, physical infrastructure, and the decisions that follow from what those systems perceive.
The work combines computer vision, multimodal AI, real-time systems, and production engineering. The aim is not only to recognise what is happening, but to make the resulting intelligence useful, timely, and safe.
Current focus
My current focus is physical AI for security. Sentinel is an early concept for adding an intelligence layer to existing CCTV systems so that footage can become searchable, important activity can trigger alerts, and incidents can be reviewed more quickly.
This is research and development in progress, not a finished or publicly launched product. I am using it to test the technical assumptions, understand real operating environments, and determine whether the idea deserves to grow into something larger.
Some current questions
- How can existing camera infrastructure gain useful intelligence without requiring complete replacement?
- How should video search, real-time detection, and event summarisation work together?
- Which tasks should run at the edge, and which belong in cloud infrastructure?
- How should privacy, security, reliability, and human oversight be designed into the system?
- What makes an alert genuinely useful to the people responsible for acting on it?
Related writing
If you work with cameras, physical infrastructure, security operations, or related research, you can write to me.