Custom applications
Web apps, customer portals, internal tools and dashboards — built around your workflow, not around the assumptions baked into a generic product.
- Portals & internal tools
- Fast, responsive interfaces
- Roles, permissions, audit trail
Software studio · Automation · Local AI
Software and AI that work for your business.
We build custom applications, automate repetitive processes and deploy AI models that run locally, on your own infrastructure. No templates, no rented platforms — just systems built around the way you already work.
- Web & internal apps
- End-to-end automation
- Locally hosted LLMs
- Integration with existing systems
Services
What we build
We start from the actual problem inside the company, not from the technology. The stack gets chosen once we understand the workflow, the volume and the people who use it every day.
Process automation
Anything done by hand, repeatedly and predictably, can be automated: moving data, generating documents, notifications, reporting, checks.
- Workflows across apps
- Document & invoice processing
- Reports delivered automatically
Local AI (on-premise)
Open-weight models installed on your server or in a private cloud. Documents and conversations stay inside the company, with no per-seat subscriptions.
- Assistant over internal docs
- Data extraction from PDFs
- Classification and summarisation
Integrations & APIs
We connect the systems that don't talk to each other today: ERP, CRM, invoicing, online store, legacy applications or files still passed around by email.
- Two-way synchronisation
- Your own documented APIs
- Connectors for legacy systems
Data & reporting
We collect data from every source, clean it and turn it into numbers you can check daily — instead of spreadsheets rebuilt by hand every week.
- Data pipelines
- Operational dashboards
- Threshold alerts
Support & evolution
We stay on after delivery: monitoring, updates, optimisation and new features, at the pace the business actually grows.
- Monitoring & backups
- Security updates
- Planned iterations
Local AI
Artificial intelligence that never sends your data away
Most companies simply cannot upload contracts, HR files or production data to a public service. The answer isn't to give up on AI — it's to bring it in-house: open-weight models running on your own server, inside a network you control.
- Your data stays with you. No uploads to third parties, no training on company information.
- Answers from your own documents. We index procedures, contracts, quotes and manuals, with citations back to the source.
- Predictable cost. Runs on hardware you own, with no per-query billing.
- Fits internal policy. Role-based access, logging, network isolation.
Internal sources
Your own server
Your team's apps
Traffic stays inside your network. Nothing leaves the perimeter.
Process
How we work
Four clear stages, with something to show at every step. You always know where we are and what comes next.
-
01
Discovery
We sit down with the people doing the work. We map the current flow, the bottlenecks and what "solved" means in numbers.
-
02
Prototype
We quickly build a working version of the highest-impact part. You test it on real data before committing to anything large.
-
03
Implementation
Development in short iterations, with regular demos, integration into existing systems and training for your team.
-
04
Operation
We monitor, update and extend. The documentation and the code remain yours, with no forced dependency on us.
Solutions
Problems we solve often
Situations we meet in manufacturing, distribution, professional services and retail.
Documents handled by hand
Invoices, delivery notes and orders read automatically, validated and passed on to accounting or the ERP.
“Where is that information?”
An internal assistant that answers from procedures, contracts and quotes, pointing back to the source document.
Apps that don't talk
Stock, prices, customers and orders kept in sync between the store, the ERP and your sales channels.
Reports rebuilt in Excel
The same numbers, generated automatically at the same hour from the source of truth — with history and alerts.
Customers calling for status
A portal with order status, documents and automatic notifications, instead of repeated calls and emails.
Requests sorted by a person
Emails and tickets classified, prioritised and routed automatically to the right team.
Technology
Tools chosen to last, not to trend
- TypeScript
- React
- Next.js
- Node.js
- Python
- FastAPI
- PostgreSQL
- Redis
- Docker
- Linux
- Ollama
- vLLM
- llama.cpp
- pgvector
- n8n
- REST & GraphQL
Frequently asked
What clients ask us
Do you work with small companies?
Yes. How clearly the problem is defined matters more than the size of the company. Many projects start with a single automated workflow and grow from there.
How long does a project take?
It depends on scope, but we work in short iterations: the first working prototype reaches you early, and from there we build incrementally, with priorities set together.
Do we own the code?
Yes. We hand over the source code, the documentation and access to the infrastructure. We don't build artificial dependencies — the collaboration continues only if it still makes sense for you.
Why local AI instead of a cloud service?
When the data is sensitive — contracts, HR files, production recipes — running locally removes the entire question of where the information ends up. Where that isn't a concern, we use cloud services too; we decide per case, not on principle.
What hardware does local AI need?
For most internal use cases, a single GPU server covers a whole team. We assess real usage volume and recommend the smallest configuration that holds up, without over-provisioning.
Do you take over projects started by someone else?
Yes. We start with a technical review of the code and infrastructure, then propose a plan: continue, partially refactor or rewrite, with the reasoning behind each option.
Contact
Tell us what's slowing you down. Let's look at it together.
Write us a few lines about the process eating your time, or about the application you need. We'll come back with a first assessment and concrete next steps.
We reply to every message, even when the conclusion is that we're not the right fit for the project.