Hours spent re-checking drawings and invoices by hand. Overcharges nobody had time to catch. Documents too sensitive to paste into a public AI. We build the private AI model, program or application that fixes your specific problem, and we run it for you.
The need
“We check every invoice against the contract by hand, and we still miss overcharges.”
What we build around it
Private by design, hosted by us or in your own tenant.
What we build
Every project starts from a specific need: who will use it, what they do today and what they want to be able to do. One team takes it from the model to the finished app.
A model trained on your documents and your terminology, to answer the questions your team actually asks, and run where you decide rather than at a third-party AI vendor.
Software built around one specific job in your business, such as checking drawings, analysing spreadsheets or protecting records, so nobody has to bend a generic tool to fit.
Mobile and web applications shaped around your users and what they want to do, for iOS, Android and the browser, designed and built end to end by one team.
Who we work with
Private AI
Most AI products are a wrapper around someone else's model. When the documents are tenders, contracts or client records, that is a risk. We build the model, and we build it to run privately. How private deployment works
Your documents
Tenders, contracts, drawings, records
Private environment
Your private model
Trained on your documents
Programs and apps
What your team uses
Hosted by us, or in your own tenant by agreement
Third-party AI vendors
Nothing is sent to them
Models run on infrastructure the client controls, whether that is our hosted environment or, by agreement, the client's own tenant.
Documents are not forwarded to OpenAI, Anthropic, Google or any other model provider. The model is our own.
Each customer's files are processed in a workspace scoped to their account, are never used for training, and can be exported or deleted on request.
The product
Screens from the Marapone construction suite, which we built with Marapone.



Client deployment
A construction-trained model, built and run by Marapone rather than rented from an AI vendor. It's the engine inside eight programs that open in a browser tab and follow the job onto site.
Marapone is a separate company. We built MaraponeAI with the Marapone team; Marapone owns and operates it.
Programs built on the model
Client deployments
Lab R&D and programs
Solutions
Models that read drawings, specifications and subcontractor quotes.
See what we buildModels that read shipping paperwork and catch problems before they cost money.
See what we buildModels that turn contracts, reports and records into decisions you can trace.
See what we buildCustom internal tools and models that remove repeated work from a team's week.
See what we buildHow we work
We start with the job you need done, who does it and what they use today, then agree what done looks like before anything is built.
One team designs, trains and builds it around those needs. You see working software early and often, not a slide deck.
We deploy it, operate it and keep improving it, so the people who built it are the people who support it.
Ways to work with us
Starting at $2,500
2 to 3 weeks
We build the smallest real version of your system and test it against your actual production data. You get a working prototype, an honest feasibility analysis and clear go or no-go criteria before you commit to a full roadmap.
Discuss thisCustom scope
Full engineering lifecycle
Production-grade architecture, private infrastructure, interface polish and complete source code transfer.
Discuss thisEnterprise
Models trained on your domain
Custom models trained on your documents and deployed into your own cloud tenant or on-premise hardware, with no third-party commercial AI API in the path.
Discuss thisTell us what you need and who it is for. We will reply by email.