DECODER · Project

AI-Powered Tools That Help Software Teams Understand and Maintain Code 20% Faster

digitalTestedTRL 5

Imagine you inherit a massive codebase at work — thousands of files, half-written documentation, and nobody around who remembers why things were built that way. DECODER built smart tools that read through all that code and documentation, connect the dots between requirements and actual code, and flag when changes in one place break something else. Think of it like a GPS for navigating complex software — it tells you where you are, what's connected, and what will happen if you change direction. The team tested it on real-world software including Linux drivers and OpenCV image processing libraries.

By the numbers

20%

Expected reduction in developer effort on medium-criticality applications

Consortium partners across the project

Countries represented in the consortium (AT, DE, ES, FR)

57%

Industry partners ratio in the consortium

Total deliverables produced by the project

Working demonstration deliverables

SMEs participating in the consortium

The business problem

What needed solving

Software teams waste enormous time trying to understand existing code, trace requirements through to implementation, and assess the impact of changes — especially on medium-criticality systems in IoT, cloud, and operating systems. When documentation is scattered, informal, or outdated, developers spend more time reading and searching than actually building. This knowledge loss compounds every time a team member leaves or a codebase grows.

The solution

What was built

The team built a suite of working prototypes: an Eclipse-based IDE with methodology support, a semantic navigation GUI that lets developers browse code by meaning rather than file structure, trace recovery tools that automatically link requirements to code, and translators that convert informal specifications into formal ACSL/ACSL++ and JML annotations for automated verification. All tools feed into a Persistent Knowledge Monitor that stores and manages extracted knowledge across the full software lifecycle.

Audience

Who needs this

Embedded software companies building safety-critical automotive or aerospace systemsCloud platform providers managing large microservice architecturesMedical device software firms needing regulatory traceabilityDefense contractors maintaining long-lifecycle codebasesEnterprise IT departments modernizing legacy systems

Business applications

Who can put this to work

Automotive Software

mid-size

Target: Embedded software companies building safety-critical vehicle systems

If you are an automotive software supplier dealing with mounting regulatory pressure to document and trace every requirement in your ADAS or powertrain code — this project developed a Persistent Knowledge Monitor and traceability tools that automatically link requirements to code and flag inconsistencies when changes are made. Tested on real codebases like Linux drivers, the tools showed an expected 20% reduction in developer effort on medium-criticality applications.

Cloud & IoT Services

enterprise

Target: Cloud platform providers managing large distributed codebases

If you are a cloud services company struggling with developer onboarding and knowledge loss when engineers leave — this project built an IDE with semantic navigation that extracts knowledge from informal documentation and code comments using NLP, turning tribal knowledge into searchable, structured information. The toolset was specifically designed for IoT and cloud computing domains with 33 deliverables covering the full development lifecycle.

Financial Technology

any

Target: Fintech firms maintaining regulated trading or payment platforms

If you are a fintech company where code changes require formal verification and audit trails — this project developed translators that convert informal specifications into formal annotations (ACSL/ACSL++ and JML), enabling automated proof and compliance checking. With 7 consortium partners including 3 SMEs bringing real industry experience, the tools bridge the gap between what regulators demand and what developers actually write.

Frequently asked

Quick answers

What would this cost to implement in our development workflow?

The project produced open-source prototypes and Eclipse-based tools, so initial licensing costs may be minimal. Integration effort would depend on your existing IDE and codebase size. Contact the coordinator through SciTransfer for current licensing terms and integration support options.

Can this scale to our codebase of millions of lines of code?

The tools were tested on substantial real-world codebases including Linux drivers and OpenCV libraries, which are large-scale open-source projects. The Persistent Knowledge Monitor was designed to handle ongoing knowledge extraction from big data sources. Based on available project data, scaling to enterprise-grade codebases was a design goal but production benchmarks are not publicly documented.

Who owns the intellectual property and can we license it?

The consortium of 7 partners across 4 countries developed the IP under Horizon 2020 RIA rules, meaning results are owned by the partners who generated them. With 3 SMEs and 4 industry partners (57% industry ratio), there is strong commercial motivation to license. SciTransfer can facilitate licensing discussions with the right consortium member.

How does this integrate with our existing development tools?

The project built Eclipse-based tool support as a primary integration point, which connects to most Java and C/C++ development environments. The GUI for semantic navigation and change assessment was designed as an interactive layer on top of existing repositories. Based on available project data, the tools work with standard version control and can process existing documentation.

What is the actual productivity gain we can expect?

The project objective states an expected average benefit of 20% in terms of developer effort, measured across use-cases in IoT, cloud computing, and operating systems domains. This was validated on real codebases including Linux drivers, OpenCV, and several Java projects (Joram, Lutece, Sat4j, Asm). Actual gains will depend on your codebase complexity and current documentation quality.

Is this ready for production use or still experimental?

The project produced 6 demonstration deliverables and 33 total deliverables including working prototypes of the IDE, semantic navigation GUI, trace recovery tools, and formal specification translators. These are tested prototypes validated on real code, not just research papers. However, they would likely need engineering work to reach full production readiness.

Consortium

Who built it

The DECODER consortium brings together 7 partners from 4 European countries (Austria, Germany, Spain, France), with a notably high industry ratio of 57% — 4 industry partners alongside 1 university and 1 research organization. Three of the partners are SMEs, which signals commercial ambition beyond pure research. The coordinator, Technikon from Austria, is itself an SME with a planning and research focus, suggesting practical orientation. This industry-heavy mix means the tools were built with real developer workflows in mind, not just academic publishing goals. For a business considering adoption, this consortium composition increases the likelihood that someone on the team is motivated to bring these tools to market.

TECHNIKON FORSCHUNGS- UND PLANUNGSGESELLSCHAFT MBHCoordinator · AT
TREE TECHNOLOGY SAparticipant · ES
COMMISSARIAT A L ENERGIE ATOMIQUE ET AUX ENERGIES ALTERNATIVESparticipant · FR
SYSGO GMBHparticipant · DE
OW2participant · FR
CAPGEMINI ESPANA SLparticipant · ES
UNIVERSITAT POLITECNICA DE VALENCIAparticipant · ES

How to reach the team

Coordinator is Technikon Forschungs- und Planungsgesellschaft MBH, an Austrian SME. SciTransfer can facilitate a direct introduction to discuss licensing or collaboration.

Order a report on this consortium See TECHNIKON FORSCHUNGS- UND PLANUNGSGESELLSCHAFT MBH profile →

Next steps

Talk to the team behind this work.

Want to explore how DECODER's code intelligence tools could reduce your development costs? SciTransfer can arrange a briefing with the right consortium partner for your use case — contact us for a personalized introduction.

Order a report on this topic More in Digital & ICT