We build software.
Some of it for clients.
Some of it for ourselves.
A software engineering and product studio. We design, build, ship and then operate what we ship, which is a different discipline from delivering it.
Most studios sell the top layer. The reason our own products still run is that we are responsible for all of it, including the parts nobody sees.
Product strategy
What to build, what to cut, and the shortest honest path to something shippable. We do this for ourselves before every product, so we are not guessing when we do it for you.
Product design
Interfaces built for the platform they live on, in the native idiom rather than a template stretched to fit. Detail that survives being looked at closely.
Native mobile engineering
Swift and SwiftUI. Offline-capable, local-first where the data is personal, and fast enough that nobody thinks about it. Four of our apps are on the App Store.
Web and edge systems
Data pipelines, APIs and applications on Cloudflare's edge. Scheduled ingestion, full-text search over millions of records, and monitoring that runs close to the thing it monitors.
Launch and operation
App Store submission, release management, uptime, incidents, and the unglamorous maintenance that decides whether software is still working in a year. This is the layer that proves the other four.
An agency that only delivers never meets the second half of the job: the failed payment at midnight, the customer who cannot get past the sign-up form, the certificate that expires on a Sunday. We meet all of it, because we own two products and we are the ones on call for them.
The most direct proof of that is unglamorous. Firme Azi runs on AL+, our own monitoring, so a fault in either one lands on us before it lands on a customer.
Firme Azi
Every company registered in Romania yesterday, on your desk this morning.
What it is
A daily feed of newly registered Romanian companies, scored and segmented for the people whose job is to call them first.
Why we built it
Romania publishes its company registry, and then leaves it effectively unusable. Anyone whose job is to reach a new business first was rebuilding the same scraper, badly.
What we had to engineer
A daily ingestion pipeline over ANAF and ONRC public registry data, normalised across all 42 counties into a searchable history of over four million companies. Each record is scored 0–100 from contact validity, sector match, days since registration and entity status. Outreach state and CSV export sit on top.
What it proves
A scheduled ingestion pipeline that has to be correct every business day, over public data nobody else has normalised, with a scoring model and a search index on top.
AL+
Uptime, cron heartbeats, errors and analytics. One toolkit instead of three.
What it is
The three services you wire up every time you ship something, collapsed into one project, one login, one bill.
Why we built it
Shipping anything small means standing up three services, three configs and three invoices, none of which agree on what a project is. We were doing it for every product we own.
What we had to engineer
An Elixir/Phoenix application on AWS, with the checks themselves running from Cloudflare's network so they originate close to the thing being checked. The BEAM is the reason one box can hold tens of thousands of concurrent checks and still recover from a failed one in isolation. Cron monitoring works by appending a single curl to the job. Analytics is cookie-free by design rather than by configuration. Firme Azi runs on it, which is how we find the faults before customers do.
What it proves
A concurrent system on the BEAM, distributed checks running from Cloudflare, alerting that has to be right at 03:00, and analytics designed to be cookie-free rather than configured to be.
Four apps on the App Store, each shipped and maintained by the same two hands that designed it.
Planning something that has to hold weight?
Tell us what you are building. You will hear back from Paul, who will be the person designing and writing it.