Research
Questions worth building prototypes for
The themes that guide our work, and the process that turns an idea into something that can be tested, trusted, and maintained.
What this page is, and is not
The themes below are research interests: directions we find worth investigating. They are not published results, completed projects, or available products. No products are listed on this website yet. We do not list publications, patents, or laboratory facilities because none are being claimed.
Interests
Research themes
Intelligence close to the sensor
How much of a model can run on a small device, and what has to move to the cloud? We are interested in the trade-offs between latency, power, privacy, and accuracy.
Reliable connected devices
Provisioning, secure updates, and graceful behaviour on poor networks. Connected products succeed or fail on operational details.
Software and hardware designed together
Firmware, apps, and services shaped by the same requirements rather than integrated at the end.
Safety and comfort in everyday environments
Practical monitoring and assistance that improves comfort or safety without demanding constant attention.
Trustworthy evaluation
Test sets, failure analysis, and measurement methods that show when a system is, and is not, good enough.
Process
From research to implementation
Select a stage to see what it involves. Each stage has explicit outputs so progress can be checked.
Frame the problem before choosing a solution
We define who the system is for, what must be true for it to be useful, and what could go wrong. Existing approaches are reviewed and the riskiest assumptions are listed.
Typical outputs
- Problem statement
- Success measures
- Risk register
Build the smallest thing that can answer a question
Each prototype exists to test one assumption: a sensor reading, a model, a protocol, an interface. We keep prototypes disposable so findings stay honest.
Typical outputs
- Focused prototype
- Measured behaviour
- Decision notes
Test against realistic conditions
Prototypes are tested against the success measures with realistic data, hardware, and network conditions. Failures are recorded and explained rather than hidden.
Typical outputs
- Test plan and results
- Failure analysis
- Go / change / stop
Turn what worked into maintainable engineering
Validated ideas become documented, tested implementations with monitoring, security, and support considered. What did not work is written down for next time.
Typical outputs
- Production-oriented design
- Documentation
- Support plan
1. Explore
Frame the problem before choosing a solution
We define who the system is for, what must be true for it to be useful, and what could go wrong. Existing approaches are reviewed and the riskiest assumptions are listed.
2. Prototype
Build the smallest thing that can answer a question
Each prototype exists to test one assumption: a sensor reading, a model, a protocol, an interface. We keep prototypes disposable so findings stay honest.
3. Validate
Test against realistic conditions
Prototypes are tested against the success measures with realistic data, hardware, and network conditions. Failures are recorded and explained rather than hidden.
4. Refine
Turn what worked into maintainable engineering
Validated ideas become documented, tested implementations with monitoring, security, and support considered. What did not work is written down for next time.