Deep sensing expertise, alongside your own team.
We build sensing instruments for a living, which means we have made most of the mistakes already. Companies bring us in for the questions that sit between disciplines, where the front end, the signal processing and the models all constrain each other and no single specialist owns the answer. We work inside your team, one or two days a week, on the decisions rather than the tasks.
Twenty-five years of measurement systems, across most of the ways a machine can sense the world.
Spectrum monitoring, direction finding, multilateration and TDOA geolocation, propagation, link budgets and antennas, in electronic warfare systems fielded in operational use. Millimetre-wave radar for navigation and detection, counter-drone sensing, and the trade-offs between range, resolution and clutter.
Acoustic sensing and ultrasound measurement, transducer behaviour, time-of-flight and echo processing, and the noise and coupling problems that dominate them.
Multispectral imaging, active illumination, global-shutter systems, radiometric calibration, and the optical design decisions that decide what a camera can actually resolve.
Detection, segmentation and classification on real imagery, running at the edge under power and latency limits, and holding up on data that is nothing like the training set.
Anomaly detection, time-series and statistics, and multi-sensor fusion: combining several imperfect sensors into a measurement none of them could make alone.
AI-assisted PCB design end to end: requirements, component selection, schematics, layout and routing. Heat management, RFI, and the sampling front ends where the analog and digital sides meet.
Enclosures, chassis and mechanisms designed for 3D printing, so a revision takes hours instead of weeks. Tolerances, stiffness, vibration, sealing and thermal paths, on machines that have to survive outdoors.
GNSS and RTK positioning down to centimetre level, autonomous driving, path planning and obstacle avoidance, on machines that operate unattended with nobody standing next to them.
Cellular and IoT links from the field, the cloud infrastructure behind them, and continuous delivery into both. Over-the-air updates, telemetry and fleet management for machines you cannot physically reach.
Underneath all of it sits the same discipline: signal processing, fixed-point arithmetic, dynamic range, calibration and error budgets. That is usually where a measurement is won or lost.
The shapes these tend to take. If one of them is familiar, we will know quickly whether we can help.
False positives and false negatives are set by the signal chain long before the model. We find where noise, dynamic range, quantisation and aliasing are being given away, quantify what each costs, and say which fixes are worth making.
A design nobody dares change is already a business risk. We recover the architecture, the requirements and the engineering intent while its author is still there, and add the testing that makes change safe afterwards.
Ambient light, temperature, vibration, dust and operators. We have taken sensing hardware from a bench into MIL-STD environments and into agricultural fields, and the failure modes are usually calibration and assumptions, not components.
Radar, optical, acoustic or a combination, and what each will really deliver at your range, resolution and price point. Deciding this late is one of the most expensive mistakes a sensing product can make.
Architecture that spans hardware, firmware, models and cloud, plus the build and release discipline underneath it. We have run air-gapped deployments and continuous delivery on the same codebase.
The algorithm that works, the test bench the line depends on, and a licence bill that grows with the team. There is a way out that keeps the physics and the instruments and loses the lock-in. How we get a company off MATLAB and LabVIEW.
Because the hardware design is not written down in a form anything can read. Documented intent, a reproducible build and automated testing are what put a hardware codebase within reach of the same assistance.
A seat at the table, not a statement of work.
One or two days a week, on site, inside the team rather than reporting to it. Long enough to understand the product properly, short enough that it stays affordable and that your engineers keep ownership of their own design.
Where a decision needs evidence rather than opinion, we will build the thing that settles it: a measurement, a model, a bench test.
We begin with a short technical session at your premises, with the engineers who know the product, to work out whether there is a problem worth our time and yours. No preparation needed and nothing to sign.
If it turns out you do not need us, that is a useful answer too, and it costs nobody anything.
Tell us what the measurement is supposed to do and where it is falling short. That is usually enough to know whether we can help.
Get in touch