Start with reality.
Physical spaces have limits. Geometry, wiring, occupancy, connectivity and human expectations shape what a system can responsibly do. We build from those constraints.
Intelligence should belong to the world we live in.
The next interface will not always be a screen. It may be a room that understands what matters, a building that makes its state legible, or a system that knows when to ask before it acts.
Give the physical world a shared language.
Today, drawings, device configurations, resident interfaces and operating systems often describe the same place differently. An agent arriving in that environment inherits the contradictions. More fluent language alone cannot repair them.
Our work begins with a shared spatial model: a way to carry the identity of a room, the capability of a device and the meaning of a decision across the whole lifecycle. It connects the work of designing a place with the experience of living in it.
Make meaning portable.
Manager, Studio, the resident experience and edge execution should be views of a coherent system. Shared identity and explicit contracts make that direction practical.
Build environments that learn.
Our longer-term research explores behaviour-centred environments: systems that translate goals into target states, learn from corrections and retain an accountable route to action.
A platform with a physical foundation.
Homira connects this ambition to the home. Tecco’s wider platform work connects authoring, engineering and operation. Research opens the next questions; product development makes them answerable in real environments.
THE COLD-START QUESTION
A home should not have to explain itself from nothing.
A generic assistant enters a room without its history. It may not know the boundary of the space, which switch controls which circuit, or the difference between a person’s preference and a device’s limit.
Design and delivery already create much of that context. Our opportunity is to preserve it: carry room geometry, engineering identity, capabilities and reviewed decisions forward into the resident experience. The model of a place can begin before the first conversation.
That is the connection between design automation and spatial intelligence. One makes the environment legible; the other explores what becomes possible when reasoning starts from that foundation.
Describe the outcome.
Compose the environment.
“Make this room right for reading” is an environmental goal. It can involve illuminance, glare, temperature, noise and the person’s preferences. Our behaviour-centred research asks how to express that target state before choosing device actions.
The language model interprets the request. A separate composition process checks capabilities and constraints. Observed state and resident corrections then close the loop. This separation is central to the architecture we are developing; it is not a claim that every part is already deployed.
Read the full architectural perspective →Comfort with context
Preferences belong to people and situations. A bright room is not always the right room. Useful intelligence must understand the activity and offer a clear way to correct it.
Energy with boundaries
Optimising a load has consequences for everyday life. Flexibility needs an understanding of the environment and the resident’s priorities, alongside dependable observations.
Trust through evidence
A confident response is only the beginning. The system needs to distinguish a proposed action, an accepted command and an observed physical result.
Your most ambitious work could live here.
Bring a problem you cannot stop thinking about. Bring something you have built. Help us make intelligence useful beyond the screen.
Explore careers