Exploring what intelligent technology can responsibly do.
This is research and prototyping, not a service catalogue. We start from a real problem, build something that can be tested, and measure whether it actually helps.
How this work runs
What this pillar is, and what it is not
The distinction matters. Plenty of companies list AI as a service. We would rather be precise about a smaller claim we can actually stand behind.
What it is
Applied research and working prototypes, chosen because engineering and security discipline is the thing that makes the result trustworthy. Every initiative starts from a problem someone actually has.
What it is not
Not a catalogue of AI services. We do not present ourselves as machine-learning specialists and we do not sell model training, data labelling or AI consulting retainers.
How it connects
The same team, and the same engineering standard, as the software and security work. The difference is the problem being explored, not a different company behind it.
InteliVis
An assistive technology initiative for people with visual impairments. It is in development, not a product you can buy today — and we would rather tell you that plainly than imply otherwise.
The challenge
Getting around an unfamiliar space when you cannot see it is one of the hardest everyday problems there is. A white cane finds obstacles, but it does not tell you what a doorway leads to, where the stairs are, or that a shop you are looking for is across the road.
Our approach
InteliVis is an exploration of whether a device can do that translation: perceive a small part of the surroundings, work out what matters in it, and convey that in a form someone can act on immediately. The hard part is not the perception model. It is the latency, the reliability, and the fact that a wrong answer in this domain is worse than no answer.
How the idea works
Concept only. This is a design direction, not a working system.
How it gets built
Define
Pin down the specific situations that matter most, and the ones where a wrong answer does harm.
Prototype
Build the smallest end-to-end version that can be pointed at a real environment and tried.
Evaluate
Test with blind and low-vision users, and measure usefulness rather than model accuracy.
Iterate
Fix what the testing exposes. Pace is set by the safety of the output, not by a launch date.
We are not publishing a timeline, because we would rather be late and honest than on time and vague. If you work in assistive technology, rehabilitation, or mobility research, we would like to hear from you.
Have a problem worth exploring?
Tell us what the problem is and who it affects. We will tell you honestly whether this is something we can help with, and what the first step would be.