Work
Systems in production, with the figures we are allowed to publish.
The work clusters in three places — financial services, telecommunications, and enterprise systems — and these four pages are the ones we can publish. More engagements are not here, because the write-ups are not finished and we would rather show you four cases we can stand behind than ten with adjectives. Every case below names what was measured, what was not, and what we are contractually unable to say.
-
Financial services · compliance
When the regulator asks why, the system answers with the sentence
For an international financial holding: regulatory text compiled into executable monitoring rules, in the language it was written in, with every live alert traceable back to the exact characters of the clause behind it. Declarative logic at runtime, not a model. Interface on Jmix.
- Built
- 2025–26
- Client
- under NDA
- Team
- 11
- Status
- in implementation
- Jmix
- Java
- Spring Boot
- Python
- sCASP
- Event Calculus
- LLM
-
Payments · mobile
Replacing the foundation under a payment app, while people were paying with it
Paytec's OPTO application was running on Xamarin a year after Microsoft stopped supporting it. We took over the codebase, migrated it to .NET MAUI without taking the service down, and then moved it onto .NET 10 LTS before the next deadline arrived. We still maintain it, and we are building the generation after it.
- Since
- 2025
- Client
- Paytec
- Team
- 2
- Status
- in production
- .NET MAUI
- C#
- .NET 10
- iOS
- Android
- PHP
- Laravel
- Vue
- SCSS
-
Telecommunications
Finding the fraud in a hundred and sixty characters
A detection engine for SMS traffic: content classification, text normalisation that survives deliberate obfuscation, campaign linking, and a labelling pipeline that turns analyst decisions into the next version of the models. Plus the Bayesian change-point work being developed on top of it.
- Built
- 2025–26
- Client
- under NDA
- Team
- 10
- Measured, our bench
- 4.1–4.3×
- PyTorch
- GPU compute
- Bayesian inference
- Text classification
-
Public sector
A city that has to answer its residents — and can prove it did
A web portal and a mobile application for a municipal administration: residents report problems with public infrastructure, staff assign and track them, crews close them with a photographed work report — and the resident who reported the problem signs it off. Built on CUBA Platform, the framework that became Jmix.
- In production since
- 2020
- Client
- under NDA
- Team
- 6
- Residents served
- ~80,000
- CUBA Platform
- Java
- PostgreSQL
- iOS
- Android
Why this page is short
Far Rainbow has been building software since 2020 and has invoiced rather more clients than the four listed above. Most of that work cannot be published: it was done under agreements that cover the customer’s name, their traffic, their internal processes, or all three. We are not going to quietly break one of those to make this page longer.
What we can do, and what all four pages above do, is publish the system instead of the customer — the workflow, the architecture, the decisions and the trade-offs, with every number either defensible or explicitly absent. If you need a reference before you can commit, say so in a meeting: there is a good deal we are allowed to tell you in private that we cannot write here.
Next
Does one of these look like your problem?
The useful conversation is not about our portfolio. It is about the thing that is breaking in your organisation and whether it is the shape we know how to fix.