Built for the hard products.
Software is getting cheap to make. Judgment is not. Implants, fleets, operating rooms, thousands of doors: we build the AI-first products where a bug is a recall, with the founders and executives who own them.
Proof
Products a business runs on.
What changed for the operator, the surgeon, the fleet, and the building, in their words and their numbers.
Selected work
They came with one hard problem. Now the product runs the business.
Wait times down a quarter. Utilization up fifteen percent.
From patient tracking to the operating system of the surgery center
Every patient tracked from registration to discharge. Scheduling, staff, and families on one platform, with the bottleneck seen before it forms.
Patients onboarded after surgery: from 30% to 95%.
Implant data turned into proactive care
Recovery predicted from six raw signals, patients reached in the first weeks after surgery, and the verification evidence that went into the client's De Novo submission.
One app for thousands of doors, whoever made them.
The software that lets every door speak to every system
From a management app for an interoperability engine to the platform Z9's ecosystem partners build on. Deployed in Fortune 500 facilities.
Staff efficiency tripled overnight.
A hardware company became a platform company
LotWatch is now the core of what Miotiv sells. The hardware is what it runs on.
In their words
“The SpiceFactory team goes above and beyond to deliver applications our patients and clinicians love. They have been a great partner in delivering significant clinical value with our smart implant technology.”
How we choose
Two to four new partnerships a year.
The rest of the year belongs to the partners who stayed. Our longest partnership began in 2015. Nobody keeps a vendor that long. They keep a partner. A partnership succeeds when the partner's business does, so we choose the ones where we can make that true.
See how we partnerHow it starts
A month in, you hold a working slice.
One conversation
What the business needs to win, who owns it on your side, what you have tried. A founder reads your message, and the first call is with a founder or our CTO.
A working slice
About a month on your real data or your real device, measured against criteria you sign off in week one. Not a deck. Not a prototype on sample data.
The plan and the price
The architecture, the evidence plan where a regulator is in the loop, and a fixed quote for what comes next.
Your decision
The slice, the architecture, and the plan are yours whether we continue or not. If you go on to the next stage, the First Build is credited against it. Every partnership that lasted began this way.
Field notes from the hard products.
E2E test maintenance • behaviour-driven testing • Playwright page objects • E2E testing • QA engineering
We Stopped Writing E2E Tests Against the UI
We did not stop driving the UI; Playwright still opens a real browser. What we stopped was letting one temporary version of the interface be the language the tests are written in.
Marko Jagodic · Nov 12, 2026 · 8 min read
Read More
flaky E2E test • test automation architecture • nondeterministic tests
Flaky E2E Tests Are an Architecture Problem
When an end-to-end test fails intermittently, most teams bump the retry count and move on. We did too, until we admitted the browser was only the last link in a chain of provider deployments, drifting data, an implicit clock, shared state and UI coupling.
Marko Jagodic · Oct 29, 2026 · 9 min read
Read More
AI Deployment • Production AI • AI ROI • Regulated Industries • Pilot to Production
The AI Pilots That Demoed Well and Never Reached Production
Most AI pilots demo well and never reach production. The CIOs who set themselves a mid-2026 deadline to prove AI value have already had to report back to their boards. Here's why pilots stall one step short of production, and how to get a single one across the line with proof it pays before the next budget review.
Milos Zikic · Oct 22, 2026 · 7 min read
Read More
Bring us the hard one.
Tell us what you are building, what the business needs to win, and who owns it on your side. A founder reads every message and answers within two working days, including when the answer is no.