Risk-Minimised Legacy Modernization: Future-Proof Your Battle-Tested Business Logic.

Our core promise: we have strong guarantees at every stage that the original functionality is preserved, because a green test suite is present at every point in time. The individual stages build on each other:

Stage 1: Classical, rule-based transpilation: A deterministic transpiler produces a semantically exact 1:1 mapping of the legacy code in the target language. This runnable reference is the safety net, no AI, no hallucination risk.

Stage 2: Test generation & trace recording: Using the runnable reference as a baseline, we record real trace logs and automatically generate a complete test pyramid. These tests approximate a “manifold” around the actual system behaviour and make that behaviour measurable.

Stage 3: Code documentation & reverse engineering: AI-assisted extraction and documentation of system architecture, business logic and dependencies: the knowledge foundation for safe refactoring and subsequent know-how transfer.

Stage 4: Custom tooling & MCP server: Project-specific tools and a context-loaded MCP server give the AI a precise semantic frame that structurally excludes code anomalies.

Stage 5: AI-based Spec-Driven Design refactoring with expert knowledge: Only now do we transform the code into a modern, idiomatic target architecture. Every transformation is validated against the test pyramid and reviewed by senior architects (Human-in-the-Loop).

Because a complete test pyramid enclosing the original behaviour is present at every point in time, we guarantee preservation of the original business logic, while delivering readable, maintainable and modularly structured code.

For banks, insurers and the public sector, the critical question about AI-assisted work is: “Where does our source code go?” Our answer: we align usage strictly with your compliance requirements. We have the expertise to cover the full spectrum; price and timeline vary depending on the level of restriction.

Operating Models: from Cloud Standard to Complete Isolation

On-premises: Full operation within your own environment is possible; we advise and accompany the corresponding setup.

European hosting with open-source models: Data processing within the EU with open-source models for maximum transparency and data sovereignty.

Dedicated, isolated infrastructure: Dedicated machines with additional protective measures, and closed-source models via controlled environments (e.g. Azure and comparable providers).

Open-weight models for highly sensitive cases: For particularly critical code, we use open-weight systems and achieve comparable quality, at correspondingly higher effort and price.

Consistent Governance Principles

Human-in-the-Loop: Every AI generation is subject to mandatory expert review.

GDPR-compliant & traceable: Transparent, controlled and deterministically reproducible AI usage.

Provider independence: Open-source practices, no proprietary lock-in. The MCP server remains with you licence-free.

No model lock-in: The approach is provider-independent: models and tools are interchangeable.

Whether public-cloud efficiency or maximally isolated processing: we have the knowledge and tools to handle any case securely. Security requirements can affect price and timeline, but never the quality of the result.

We deliver large-scale projects in heavily regulated core domains. Our successful engagements include:

Société Générale (Host & Test Automation): Cloud migration of a business-critical core banking system from a grown mainframe architecture. Volume: 10 million lines of code, 17,000 programs, 12,000 batch jobs. Specialised tools achieved an automation rate of 99.99 %. Functional equivalence was verified error-free via automated production traces (Capture & Replay) in real-time comparison; hundreds of person-days in testing were saved in the process.

Zurich Group (Legacy Migration & DSL Build): Structural consolidation and highly automated migration of approximately 200,000 lines of historically grown mathematical C code for life insurance into a domain-specific language (DSL). The previously up-to-four-week manual programming phase per release was completely eliminated; an integrated code generator produces immediately testable code.

APG Affichage (Oracle Forms → Java): Complete migration of a historically grown Oracle Forms application into a modern Java architecture including project management, legacy system analysis and DSL-based migration methods. The user interface was automatically and error-free regenerated via code generation. Stack: Java, Spring, Oracle RDBMS, JPA, JSF, Eclipse GMF, Xtext.

Large Life Insurance Legacy System (Know-How Preservation): Model-based, highly automated migration to drastically reduce maintenance costs and safeguard against demographic change in the IT team. Extremely short freeze time, parallel development of legacy and new system, 100 % freely definable target architecture.

International Reinsurer (Mainframe Replacement): Technical project coordination for the migration of a complex host claims calculation application (commission and settlement logic) to a modern Java platform, including algorithm optimisation, a modern testing strategy and a quality management manual.

Deutsche Bahn & parcIT (AI-Assisted Toolchain): Building an AI-assisted toolchain integrated directly into existing development processes, including MCP servers, IDE integrations, an automated review bot anchored in JIRA/Confluence and a business analyst tool. Evidence of controlled, deterministic and traceable AI usage including change management and know-how transfer.

Generali (since 2019): Long-term partnership covering architecture consulting, language engineering and continuous performance optimisation of business-critical systems.