What we do

Product development under one roof

Most hardware programmes fail at the seams. A design consultancy draws a beautiful enclosure, an electronics contractor lays out a board to fit a volume that has since changed, and a manufacturer receives a package that nobody has ever built. The client sits in the middle translating between three groups who have never spoken to each other, and the mistakes surface in month five when they are expensive.

We take on products where the hard part is getting several disciplines to agree at once. On a small device the shell, the board, and the code are not three problems, they are one problem viewed from three angles. A millimetre that moves in the enclosure is a board respin. A sensor that misbehaves at the edge of its range is a firmware problem and a mechanical problem at the same time. Keeping those decisions in one room is the whole point of the studio.

Industrial design

Form, use, and how an object is held, worn, or mounted. We draw the shell around a real stack of components and a real hand, then keep redrawing as both change. A shape that only works in a render is the most expensive kind of pretty.

Ergonomics, materials, finish, and the small courtesies of use get resolved alongside the engineering instead of negotiated against it at the end. Where a button lands, how a lid closes, what the device does when you set it down. Those are decided while there is still room to decide them.

Mechanical engineering

Enclosures, mechanisms, materials, tolerances, and design for assembly. Sealing and ingress protection where a device has to survive a clinic or a pocket. Thermal paths where something inside runs hot and the outside has to stay comfortable against skin.

Tolerances, materials, and assembly order arrive with the drawings rather than after them, so the first build tells you something true instead of something hopeful.

Electronics and PCB design

Schematic and board design, power, sensing, and wireless. We work at the scale where the board is smaller than a thumbnail: rigid-flex on thin polyimide, fine-pitch BGAs, and devices that run for weeks on a battery you would struggle to feel. Budgets like that are won in nanoamps of quiescent current and milliseconds of wake time, and an antenna has to be matched against a body rather than against a bench.

Mixed-signal instruments are the work we are proudest of, where microvolt analog, sensitive optical detection, high-voltage actuation, switching supplies, and controlled heat all have to share one enclosure without ruining each other. Making hostile subsystems coexist in a small volume is the thing we are best at.

Biopotential acquisition at the microvolt level, where the design problem is settling, crosstalk, and noise across many electrode channels rather than the choice of converter. Precision thermal and fluidic control, with sensor arrays resolving hundredths of a degree, closed-loop pressure regulation, and valve manifolds. Wireless across several silicon families, including Bluetooth Low Energy, LE Audio broadcast to a room full of receivers at once, and millimetre-wave radar for contactless sensing.

Firmware and embedded software

The behaviour of the device itself, written beside the board it runs on by people who can walk over and put a probe on it. Bring-up starts while the board is still on the bench, which is where power draw, timing, and the awkward edges of a sensor actually show up. Firmware written after a design is frozen inherits every compromise in it.

Embedded C on Cortex-M class parts. Layered architectures that port between microcontroller targets, interrupt-gated acquisition, double-buffered streaming, and fixed-point signal processing where there is no floating point to spare. Power engineering as a first-class activity rather than a late optimisation: single-digit microamp sleep, peripheral teardown, wake on sensor interrupt, and per-unit production tests that measure it instead of assuming it.

On-device machine learning inside real constraints, with convolutional models a few tens of kilobytes across running on parts that have a couple of hundred kilobytes of flash, detecting an event in under a tenth of a second, and updated over the air with validation and rollback instead of a recall. Where the product warrants it we work to MISRA-C, sub-millisecond worst-case task periods, and fault recovery budgets, because on some instruments a stalled loop is a safety problem.

Software around the device

Companion software that does real work: desktop and web hosts for instrument control, live telemetry, calibration, firmware update, and the data pipeline behind a sensor. Device to cloud where the product needs it, with mutual TLS and store-and-forward for connections that are not reliable.

Machine learning pipelines from capture and labelling through to a model small enough to run on the device itself. We also run our own operations on software we built, which is part of why we can be specific about what maintaining one actually costs.

Prototyping and in-house assembly

We assemble our own boards, including fine-pitch parts on thin flex, with optical inspection and purpose-built programming and test jigs. Keeping that in-house is what makes an iteration a matter of days rather than a month of waiting in someone else's queue.

We build prototypes to answer questions rather than to decorate a deck. A prototype should aspire to be the product rather than a throwaway lie, because a physical unit that proves a mechanism is worth more than a render that proves nothing, and finding out early costs a fraction of finding out at tooling.

Some of that speed comes from tools we built for ourselves. A modular hardware platform of interchangeable compute and low-power radio cores on a shared mezzanine bus, with a library of sensor modules behind it, designed to ship inside the product rather than be discarded after the prototype. And a debug probe that switches between I2C, SPI, UART, SWD, GPIO, and a background logic analyser on one reconfigurable ribbon, taking a new part from datasheet to a verified streaming driver in about a minute rather than an afternoon.

Design for manufacturing and production transfer

A design is finished when someone who has never met you can build it twice and get the same product. That takes drawings, a bill of materials that survives a real quote, and test steps that catch the failures you already know about.

First-pass yield is treated as a design problem rather than a factory problem. We have taken a thin-flex assembly process from a yield we were not willing to live with to comfortably above ninety per cent, and that was process engineering rather than luck. The same attention goes into assembly order, fixturing, and the tolerances that decide whether a build repeats.

We produce the documentation, coordinate suppliers, and stay through first articles and the questions that come back from the floor. That is where a design either holds or quietly gets changed on your behalf.

Medical devices and regulated work

We work on medical-adjacent hardware and design with regulated development in mind: traceable requirements, design history, risk registers, and per-unit acceptance criteria where the product calls for them.

We are candid about the line between engineering and certification. Wireframe does not issue regulatory approvals and we do not claim certifications we do not hold. Where a submission is involved we build the engineering evidence it needs and work alongside your regulatory lead or a specialist consultancy rather than replacing them. If your programme needs a certified quality system as a contractual condition, ask us directly and we will tell you plainly whether we can meet it.

The kinds of products we take on

What decides whether a brief is for us is not the sector, it is whether the hard part spans disciplines. If the mechanical, the electronic, and the software problems are tangled up in each other, that is our kind of problem.

In practice the work clusters around devices that have to survive contact with real use. Body-worn sensing, including temperature, motion, and optical vitals on devices that have to last on a body without being noticed. Point-of-care diagnostics combining electrochemistry, fluorescence, spectral measurement, and in-cartridge imaging, with the fluidics and thermal control to drive them. Neuro and biopotential instrumentation. Laboratory automation with imaging and fluid handling for equipment that runs unattended. Accessibility and assistive devices, including spatial sensing with haptic feedback and multi-channel wireless audio for live translation. Connected consumer and industrial hardware where the sensing is the hard part.

Medical wearables and diagnostics are where we have gone deepest, because they forgive the least. The skill we carry between projects is the integration rather than the subject matter, and it travels. If a brief is mostly a website, a mobile app with no hardware, or a pure research study, we are usually not the right studio and will say so.

How a project runs

Work is broken into phases so the risk is retired in order. First the brief and feasibility: what the product must do, the constraints that actually bind, and the unknowns worth testing early. Then concept and architecture, with a small number of directions carried far enough to compare honestly. Then detailed design, taking the mechanical, electrical, and software work to a buildable state. Then prototype and validation, with physical units measured against the requirements and iterated. Then transfer, meaning the documentation and supplier work that lets someone build it more than once.

Duration depends on the phase and the unknowns. We scope each phase with you rather than quoting a single number for an undefined project.