Custom AI solutions for businesses

We put AI into production — securely.

UWS develops, integrates, and operates custom AI solutions inside your existing systems — and keeps them running.

Book an Initial Call Describe Your Project

For companies with 50 to 500 employees.

From idea to production

01
Starting point
02
Approach
03
Delivery
04
Operations

Integrated into your existing systems
On premises or European cloud
Operations and ongoing development on request

Since 2010
Development, integration, and operations
European engineering team

Experience from long-term software and digital projects

01Starting points

AI potential is rarely the hurdle. Getting it into production is.

Many companies can see where AI would help. What is missing is a dependable path from the idea to an integrated solution that can be operated long term. Existing systems, individual processes, security requirements, and limited internal capacity make delivery demanding.

01
No clear entry point
You want to use AI but do not yet know which use case makes sense first.
02
Off-the-shelf tools do not fit
Your processes and systems need a custom solution, not another isolated tool.
03
Existing systems must be part of it
AI should not sit next to your system landscape — it should work inside it.
04
Operations must stay controllable
You need clarity on where and how the solution is operated.

This is exactly where UWS starts.

02Solution portfolio

AI for knowledge, processes, and custom applications

We combine the modules of the UWS AI Platform with custom software engineering. The result: solutions built on proven components that still fit your processes, systems, and security requirements.

UWS AI Knowledge

Put company knowledge to work — securely

Documents, data sources, and existing applications are connected into one controlled knowledge access layer. Employees get context-aware answers and find relevant information faster.

Typical use cases

Internal knowledge assistant, research, service and support enablement

UWS AI Agents

Automate multi-step workflows with clear controls

AI agents gather information, work through tasks across multiple systems, and involve people wherever review, approval, or a decision has to stay human.

Typical use cases

Document processing, research and reporting, recurring service and back-office processes

Custom AI solutions

Integrate AI precisely into applications and processes

UWS builds custom AI applications, integrations, and technical components for specific domain requirements. Depending on the initiative, they are created entirely from scratch or combine UWS AI Knowledge and UWS AI Agents into one complete solution.

Typical use cases

Domain applications, AI copilots, system integrations, data-intensive workflows

You stay in control of your solution — including the model, provider, and operating environment.

The UWS AI Platform is built on established open-source components and open interfaces, so models, infrastructure, and operating environments remain adaptable. That reduces long-term vendor lock-in and protects your investment.

The platform modules can be used independently, combined with each other, and extended with custom components.

Built on open source · flexible choice of provider and operating model
Integrated into your existing systems
On premises or European cloud
Operations and ongoing development on request

Discuss your approach

03Approach

How we bring solutions into production

Step 1
Understand the starting point
We clarify the task, the processes, the data, and the existing systems.
Step 2
Design the solution
We define the target state, architecture, security requirements, and the most sensible entry point.
Step 3
Build and integrate
We implement the solution, connect systems and data sources, and validate against agreed criteria.
Step 4
Operate and evolve
We support go-live and, on request, take over monitoring, maintenance, and continuous improvement.

The entry point depends on the initiative: technical analysis, pilot project, or direct implementation.

04Operations and security

Security is part of the architecture.

Architecture and operations are aligned with each customer's data protection, compliance, and security requirements. This includes GDPR and, where applicable, NIS2, the EU AI Act, and the Cyber Resilience Act (CRA). Operating location, infrastructure, and safeguards are defined jointly for each project.

Operating location

Operated in your environment

On premises

The AI solution runs inside your own infrastructure and is connected to your existing internal systems.

Operated in Europe

European infrastructure

Operated on the infrastructure of an ISO 27001 certified cloud provider in Europe, scaled and supported per project.

AI security across the entire data flow

Security does not end with hosting. We look at the full processing chain — from development and test data through project team access to production model inference.

Controlled model inference

Model and infrastructure are chosen to match your security requirements: external providers, private endpoints, or models inside your own environment. Customer data is used for training neither by UWS nor by the model providers involved.

Protected development

Development, test, and production environments stay separated. Real data is anonymised or pseudonymised, and we use AI coding tools only in approved environments where customer data is not used for model training.

Controlled access

Access to data, code, and infrastructure is granted per role following least privilege, and is logged. In systems operated by UWS, data is encrypted in transit and at rest.

Defined data lifecycles

Retention and deletion periods for prompts, logs, and test data are defined per project and enforced technically.

Transparent data flows

Model providers, processing region, and retention periods are documented before the project starts. It stays traceable which systems receive which data.

We control not only where an AI solution runs, but also which data a model processes — and under which conditions.

05Case studies

Experience that can be substantiated

Four selected projects showing how we build, modernise, and operate data-intensive systems over years. Further references on request.

Current case studyLegal tech / professional services

Building a secure AI knowledge and workflow platform

Productising a RAG MVP
Multi-tenant SaaS on AWS EU
Q4 2025 – ongoing

Knowledge sits scattered across documents, email, and line-of-business systems. We are building a secure AI knowledge and workflow platform on top of it: retrieval-augmented generation, semantic search, and agentic capabilities. A local prototype became a tenant-isolated multi-tenant base in the AWS EU cloud — development is ongoing.

Starting situation: Document-intensive organisations spread relevant knowledge across files, email, line-of-business applications, and personal storage. The information exists, but finding, assessing, and reusing it takes considerable time — and implicit knowledge is barely available during handovers, holidays, or staff changes.
Business goal: Make existing knowledge productively usable without forcing new data structures or additional maintenance work — while meeting demanding requirements for data protection, tenant separation, auditability, and controlled data access.
UWS solution: Development of an AI knowledge and workflow copilot: company data is opened up through retrieval-augmented generation, semantic search, and agentic capabilities. The MVP focuses on fast document and case analysis, contextual questions, summaries, and reuse of the results; an action and agent layer extends the system from knowledge retrieval towards productive workflows.
Data sources and integrations: Connected or planned for the MVP: PDFs and office documents, email and scans, existing file and folder structures, Google Drive, and further company sources via connectors. The platform is API-first and modular, deliberately separating infrastructure, authentication, and multi-tenancy from retrieval, parser, connector, and agent logic so AI components can evolve or be replaced independently.
Security model: The original single-user prototype architecture was developed into a multi-tenant, enterprise-ready platform: PostgreSQL with tenant isolation, row level security, JWT-based authentication, an API gateway, central logging, containerised services, a restrictive super-admin model without routine access to customer documents, and operation in the AWS EU cloud. Data protection and technical tenant separation are therefore addressed at architecture and database level.
Current status: The path led from an existing local RAG prototype through architecture and security refactoring to a cloud-ready multi-tenant product base. In early 2026 the transition from experimental prototype to a productisable SaaS system was completed; development is ongoing.

Delivery is AI-assisted, with human control over architecture, security, and releases. Further regulatory assurance for professionals bound by confidentiality duties (GDPR, §203 StGB, §43e BRAO) is a separate, ongoing compliance workstream and is not presented as complete.

Case study 02Finance and operations software

Future-proofing a business-critical finance and operations platform

Enterprise platform modernisation
4 interconnected codebases
AI-assisted delivery with human review

A business-critical finance and operations platform of four interconnected codebases ran on outdated technology. We modernised backend, frontend, security, and deployment in controlled, reversible stages — AI-assisted, with human review. The result: a validated release candidate without blocking findings and a support horizon of more than three years.

Starting situation: A business-critical platform connects finance and operational workflows across a web application, backend services, accounting integration, file transfers, and reporting. Outdated technologies and a legacy deployment model increased security, support, and release risk.
Business goal: A secure and maintainable foundation for the next phase of the product — while fully preserving the established business processes and integrations. A time horizon was defined upfront: after the upgrade, the application as a whole should have a support horizon of more than three years until end-of-life.
Modernisation by UWS: Backend, frontend, security, and deployment were moved to a current baseline in controlled, reversible stages. Compatibility and validation gates protected each step; a dedicated read-only mode safeguards business data and external integrations during migration and maintenance.
Controlled use of AI: A project-specific context and task system supported system inventory, compatibility analysis, migration planning, cross-repository review, implementation, and test preparation. Every change remained traceable and under human control.
Result: The technical modernisation and client validation of the release candidate were completed without blocking findings; production rollout is prepared. Four out of four modules pass integrated validation, 270 database changesets were verified with no pending migrations, 13 credential-bearing values were moved into deployment secrets, and the final npm audit reported no known vulnerabilities in the frontend dependencies.

AI was embedded as a controlled engineering copilot. UWS retained ownership of architecture, security, business behaviour, and release decisions.

Case study 03Software as a service

SaaS platform for workforce management

Software development fully switched to AI-driven processes
SaaS product
36 months in, development ongoing
Dedicated remote team: architecture, backend, frontend, DevOps, QA

A product that was no longer economically viable was rebuilt as a cloud-based SaaS platform — web and mobile, delivered by a dedicated remote team. Development and operations from one team for 36 months; software development has since fully switched to AI-driven processes, with human sign-off before every release.

Starting situation: The existing solution was no longer economically viable — outdated architecture and technologies, high development cost, and a shortage of developers — while user numbers and feature demand grew sharply.
Business goal: A rebuilt product that can be developed further long term, operated economically, and rolled out to additional markets.
Technical challenge: Rebuilding web and mobile applications in parallel on a cloud architecture that scales with user numbers and stays maintainable for years.
UWS solution: Concept, design, rebuild, and operation of the web and mobile applications in a cloud environment: microservices, container orchestration, and centralised logging, built on mature open-source frameworks chosen for the long run.
Integration and operations: High test coverage across all features with unit, integration, and acceptance tests. The full management and engineering team was provided as a dedicated, stable remote team.
Result: The first product version shipped within the planned budget and timeline. The product is being rolled out to further countries and is in use at UWS itself.
Current status: Development is ongoing with the same dedicated team. The engineering processes have been moved to AI-driven ways of working: AI-assisted implementation, review, and test preparation within defined guardrails, with human approval before every release.

This case shows the operating model we also offer for AI solutions: development, integration, and operations from a single team, sustained over years — and the same team now develops this product with AI-driven engineering processes.

Case study 04Business journalism

Real-time editorial system for business news

Journalism
10 months
On-site project lead, 3 remote engineers

An editorial system for business news that could no longer be evolved was replaced within ten months by a modular real-time solution: consolidating many high-frequency data sources, processing them automatically by rule, and running reliably for years — project lead on-site, development remote.

Starting situation: An existing editorial system for business news was not modularised and could no longer be developed further.
Business goal: A system that meets internal performance requirements and supports the company's strategic direction.
Technical challenge: Consolidating a large number of incoming news channels at high frequency while staying consistently stable and scalable.
UWS solution: Architecture review, modularisation, and a rebuild as a rich internet application with server push. Lead editors got a configurable rule system that processes incoming and outgoing news automatically, plus filters, archiving, and text generators.
Integration and operations: Consolidation of many news sources in near real time, backed by unit, integration, and functional test coverage and continuous delivery.
Result: Higher editorial efficiency, with parts of the news processing running fully automatically.

This project predates today's generation of AI. The task is the same: consolidate many data sources at high frequency, process them automatically by rules, and keep them running reliably.

06About UWS

One European team. Direct collaboration.

UWS is a European team for AI and software solutions. Clients work directly with experienced project leads and technical specialists. Around 25 specialists in software engineering, project management, DevOps, and quality assurance form the core UWS team, with engineering teams in Germany and Poland and company roots in Dublin. Development, integration, and operations come from one team.

2010 · Dublin, Ireland

UWS was founded in Dublin in 2010, when Sören was working at IBM Ireland and Felix at Logica. That experience in international technology companies still shapes how we develop, integrate, and operate solutions today.

Felix Schendel and Sören Kunst – founders of UWS — Felix Schendel
Founder · then at Logica
Sören Kunst
Founder · then at IBM Ireland

We believe in technology that works in the real world: close to the task, engineered with care, and dependable in operation.

Sören & Felix · Founders of UWS

01Around 25 specialists in the permanent team
02Engineering capability in Germany and Poland
03Development, integration, and operations from one team
04In business since 2010

Which task should AI take on in your company?

Describe your starting point in a few sentences. In the initial call we clarify your starting point and outline a possible approach.

Book an Initial Call Describe Your Project

We put AI into production — securely.

AI potential is rarely the hurdle. Getting it into production is.

No clear entry point

Off-the-shelf tools do not fit

Existing systems must be part of it

Operations must stay controllable

AI for knowledge, processes, and custom applications

UWS AI Knowledge

UWS AI Agents

Custom AI solutions

How we bring solutions into production

Understand the starting point

Design the solution

Build and integrate

Operate and evolve

Security is part of the architecture.

Operated in your environment

Operated in Europe

AI security across the entire data flow

Controlled model inference

Protected development

Controlled access

Defined data lifecycles

Transparent data flows

Experience that can be substantiated

Building a secure AI knowledge and workflow platform

Future-proofing a business-critical finance and operations platform

SaaS platform for workforce management

Real-time editorial system for business news

One European team. Direct collaboration.

Which task should AI take on in your company?