Hardware for Startups
Most hardware startups don't fail at the idea. They fail in the gap between a prototype that works on a bench and a product that can be built a thousand times — and that gap is where we work.
What a hardware partner actually does for a startup
A software startup can ship on a Friday and fix it on Monday. Hardware doesn't work that way. Once a tool is cut, a reel is bought, or a thousand boards are panelised, mistakes stop being edits and start being money. The job of an engineering partner is not to write firmware faster than you could — it's to make sure the decisions that are expensive to reverse get made correctly the first time, and to make the rest boring.
We work with founders from a sketch on a napkin through to parts arriving on a pallet. Sometimes that means the whole product: electronics, firmware, enclosure, app and the transfer to a contract manufacturer. Sometimes it means one subsystem, or an honest second opinion on a design someone else built. Both are real engagements, and we would rather tell you which one you need than sell you the larger one.
The five places hardware startups lose money
These are not hypotheticals. Every one of them is something we have been called in to repair after the fact, and every one of them is cheaper to avoid than to fix.
1. Choosing the platform before understanding the product
The first real decision is which microcontroller family, which radio, and whether there's an operating system at all. It gets made in week one, usually on the basis of whichever dev kit was already on the desk, and it constrains everything afterwards: power budget, certification path, firmware architecture, even which engineers you can hire. Switching an ESP32 product to an STM32 eighteen months later is not a port — it's a rewrite with a new BOM and a new certification. We spend real time on this choice before any schematic exists, because it's the one that compounds.
2. The dev kit that can never become a product
Modules and dev boards are the right way to prove an idea and the wrong way to ship one. They carry cost you don't need, a footprint you can't fit, and parts you don't control. But the transition to a custom board is also where founders get surprised, because it isn't only layout — it's component selection with availability in mind, a power tree that survives real loads, test points for a production line, and firmware that no longer assumes a USB cable is attached. Knowing when to make that jump is worth more than making it early.
3. The prototype nobody can manufacture
A prototype proves the product can exist. It does not prove it can be built repeatably by people who have never met you. Design for manufacture discovered at the contract manufacturer instead of at the design stage is the single most expensive category of surprise in hardware: tolerances that only worked because a part was hand-fitted, an assembly order that requires three hands, a snap-fit that passes at prototype and cracks at volume. We review for this while the design is still cheap to change.
4. Certification and power, discovered late
FCC and CE outcomes are decided by ground planes, filtering and board edges — which is to say, they are decided during layout, months before anyone books a test lab. The same is true of battery life, which is set by the sleep architecture rather than by the cell you choose — you can sanity-check yours in two minutes with our free Battery Life Calculator — and of shipping, where lithium cells bring UN38.3 testing and paperwork that can hold a launch. None of this is hard if it's planned. All of it is brutal if it's found at the end.
5. Not knowing what any of it should cost
Founders raising their first round are routinely quoted numbers ten times apart for the same scope, with no way to tell which is honest. So here are our real figures, from delivered work, with client names removed:
- A campaign-ready BLE tracker prototype — architecture locked, module-based first revision, printed enclosure, BOM costed at volume: 48 engineering hours.
- A Raspberry Pi 5 GPIO adapter board, schematic through released production files: about 55 hours.
- A wireless alert-light prototype, hardware and firmware together: about 85 hours.
- Firmware and system integration for an embedded-Linux audio device with on-device inference and over-the-air updates, a genuinely hard project: about 120 hours.
Most projects we quote total between 40 and 250 engineering hours. A total above 400 means a genuinely complex, regulated or multi-device programme, and above 800 is rare. If a quote you're holding is far outside those bands, it's worth asking what's in it. We publish these numbers because a founder who can't calibrate can't negotiate, and we would rather compete on being right than on being opaque.
A real one: ELDA BrainTech
ELDA BrainTech is bringing a neurodiagnostic medical device toward the FDA. We built the software behind it — the real-time acquisition path from their FPGA-based hardware, and the interfaces their team works in every day. The stack is stable, and the exploratory phase is over. The full story, including what makes medical-device software a different job, is in the case study: see Case Study: The Software Behind ELDA BrainTech's Neurodiagnostic Device.
What you own
You own all of it. Schematics, layout sources, Gerbers, drill and pick-and-place data, a costed bill of materials with alternates for anything single-sourced, firmware sources, CAD, and the bring-up test plan. You can take that package to any manufacturer on earth without asking us. We never hold design files, we never take equity, and a mutual NDA is signed before any technical detail is exchanged.
How a first engagement starts
Not with a long statement of work. We start with a short, bounded Feasibility Review, quoted up front: about a week, a written assessment of the highest-risk interfaces in your design, a recommended order of experiments that would kill the idea fastest if it's wrong, and a written quote with phases for the next stage. If you proceed, the fee is credited in full against that stage.
It exists because the honest answer to "what will this cost" is usually "less than you fear, and the risky part isn't where you think it is" — and proving that cheaply is better for both of us than a proposal built on guesses. If the review says the idea doesn't work, you've found that out for the cost of a one-week review instead of a funding round.
Frequently Asked Questions
We only have a sketch and no engineer. Is that too early to talk?
No, and it's often the best time. The decisions that cost the most to reverse — platform, radio, power architecture, certification path — are made before any hardware exists, and those are exactly the ones a founder without an engineer tends to make by accident. A 30-minute call at the sketch stage regularly saves more money than anything we do later.
What does it cost to get from an idea to a working prototype?
For most products, 40 to 250 engineering hours covers scoping through a working prototype, depending on how many disciplines it touches. A single-board project with existing firmware sits at the bottom of that; a full product with custom electronics, an enclosure, an app and a production path sits at the top. We price in engineering hours; you get a written quote with phases before any work starts. Our free estimator gives you a planning number in a couple of minutes.
Who owns the intellectual property?
You do, completely. Every design file, source file and document is yours, and you keep 100% of the IP and all project data. We never take equity, and we never make a client dependent on us to manufacture their own product. A mutual NDA is signed before any technical detail is exchanged.
Can you take it all the way to production, or only to prototype?
All the way. Prototype is the halfway point, not the finish line — the work after it is design for manufacture, tooling, supplier vetting, test fixtures and the production ramp. We do that unglamorous part too, including the transfer to a contract manufacturer, and we're happy to hand over to a CM you already have.
We have a design someone else built. Can you review it instead of starting over?
Yes, and it's usually the cheaper path. Send schematics, Gerbers or firmware and we come back with a plain list of what's actually wrong, ranked by what it costs to fix. Sometimes the answer is a small respin; sometimes it's a single component swap that solves an availability problem without touching the layout. Occasionally the answer is that the design is fine and the problem is somewhere else, and we'll tell you that too.
How fast can we start, and how do we begin?
Usually within a week. The first step is a free 30-minute technical call — not a sales call; bring the hard question you don't have an answer to. From there, most founders start with the Feasibility Review, quoted up front, so the first real spend is bounded and the next quote is based on engineering rather than guesswork.
Ready to talk specifics?
A 30-minute call with a senior engineer beats three rounds of email. Tell us what you're building.