TASKING, Infineon and DLR Win AWS Hackathon With AI-Driven Approach to Automotive Safety-Critical Software
A cross-organizational team from TASKING, Infineon and the German Aerospace Center (DLR) has earned the top award at the AWS “Accelerating the V-Cycle with Agentic AI” hackathon, highlighting a new approach to software development for increasingly complex, software-defined vehicles.
The winning team developed a working prototype designed to demonstrate how governed artificial intelligence can coordinate critical engineering activities throughout the automotive software development lifecycle. The project received the highest score in the competition for OEM purchase readiness, signaling the potential commercial value of an AI-assisted engineering workflow that connects semiconductor information, software implementation, compliance, testing, verification and evidence generation.
The achievement comes at a time when automotive manufacturers and suppliers are under increasing pressure to develop sophisticated vehicle software faster while maintaining strict safety, quality and regulatory requirements. Modern vehicles increasingly depend on software for functions ranging from powertrain management and advanced driver assistance to connectivity, vehicle control and automated driving.
As software becomes more central to vehicle architecture, development teams must manage enormous quantities of requirements, hardware documentation, source code, test cases and verification data. The challenge is not simply producing software quickly. Engineers must also demonstrate that the software satisfies requirements, follows applicable coding standards, operates correctly on the intended hardware and produces sufficient evidence for safety-critical development processes.
The TASKING-led project seeks to address this challenge by combining AI-based orchestration with trusted engineering information and deterministic verification technologies.
Addressing the Complexity of Software-Defined Vehicles
The transition toward software-defined vehicles is changing how automotive companies design, develop and maintain vehicles. Instead of vehicle functionality being determined primarily by mechanical and electrical components, software increasingly determines how vehicles behave and how new features can be introduced.
This shift brings significant opportunities but also creates new engineering challenges.
Automotive development organizations frequently operate across multiple teams, suppliers and technology platforms. Requirements may be managed in one system, semiconductor information in another, source code in development environments, testing through specialized tools and compliance documentation through separate processes.
These disconnected workflows can create manual handoffs and delays. An engineer may need to move information between multiple systems before being able to determine whether a software change satisfies requirements or complies with applicable coding standards.
The problem becomes even more complicated when semiconductor devices are updated. Product specifications, errata and other hardware information can influence how software should be implemented and verified. If that information is not incorporated early enough, downstream development and testing may be based on outdated assumptions.
The winning prototype demonstrates how AI agents can help coordinate these activities within a more connected workflow.
Rather than allowing AI to make unrestricted engineering decisions, the project focuses on a governed model in which AI assists engineers while deterministic tools and human approval remain essential parts of the process.
Combining AI With Engineering Governance
One of the central ideas behind the project is that AI alone is not sufficient for safety-critical software development.
Generative AI can produce code, documentation and tests at high speed, but safety-critical industries require much more than code generation. Engineering teams need reliable technical information, repeatable verification processes, traceable evidence and human oversight.
The prototype therefore places AI inside a structured engineering framework.
AI agents are used to coordinate different stages of development, while established engineering tools perform specialized analysis and verification. This creates a division of responsibilities in which AI can automate repetitive coordination and reasoning tasks, while deterministic tools provide objective technical results.
The approach is intended to help engineers reduce manual work without removing human responsibility from critical decisions.
According to Travis Bone, Principal Solutions Architect at TASKING, the team developed the prototype to demonstrate how AI can orchestrate engineering activities while preserving the rigor and human judgment required for safety-critical development.
The project combines TASKING’s compliance and verification capabilities with trusted semiconductor knowledge from Infineon. It also incorporates expertise from DLR, creating a cross-organizational development effort focused on the challenges of safety-critical engineering.
AWS Kiro AI Agents at the Center of the Prototype
The hackathon prototype uses AWS Kiro AI agents to coordinate activities across the software development lifecycle.
The workflow connects several engineering functions that traditionally may be handled separately. These include semiconductor specifications, product updates and errata, software implementation, coding-standard analysis, automated unit-test generation, virtual ECU execution, structural coverage analysis and evidence reporting.
This integration allows information from one stage of the engineering process to influence activities later in the workflow.
For example, trusted semiconductor information can provide the hardware context needed during implementation. Once software is developed, TASKING technologies can evaluate compliance with coding standards. Automated unit tests can then be generated and executed, while coverage analysis helps determine whether the software has been sufficiently exercised.
The AI agents can evaluate the results and initiate additional iterations when necessary.
This creates a feedback loop rather than a simple linear development process.
If coding-standard issues are identified, the workflow can help address them. If test coverage is insufficient, additional tests can be generated and evaluated. If requirements or structural coverage remain unresolved, the workflow can continue the analysis or document the remaining issues.
The goal is to reduce the amount of manual coordination required between individual engineering activities.
Connecting Trusted Semiconductor Knowledge With Software
A particularly important aspect of the project is the role of semiconductor information.
Infineon contributed trusted device specifications, product updates and errata to the workflow. These resources provide the AI-assisted engineering process with information about the actual hardware environment for which software is being developed.
This is significant because AI-generated recommendations are only as reliable as the information used to support them.
In a safety-critical automotive environment, generic knowledge about software development is not enough. Engineers must understand the characteristics and limitations of the specific microcontrollers, processors and other semiconductor components used in a vehicle.
By connecting semiconductor knowledge directly to the workflow, the project aims to ensure that software development activities remain aligned with the intended hardware configuration.
Simon Achatz, Principal Engineer and Team Lead AI Systems Automotive at Infineon, emphasized the importance of connecting implementation, verification and compliance with trusted engineering knowledge.
The approach is designed to help ensure that AI-assisted decisions are grounded in real hardware information rather than disconnected or outdated assumptions.
TASKING AI Framework Enables Continuous Verification
TASKING’s technology plays a key role in the verification and compliance side of the workflow.
The TASKING AI Framework allows agents to support activities such as coding-standard analysis, unit-test generation and coverage evaluation. Instead of treating verification as a final step after software development, the prototype demonstrates a more continuous approach.
This means that compliance and verification can occur iteratively during development.
When a potential issue is identified, the workflow can respond by refining the software or generating additional tests. The results can then be evaluated again.
Such an approach has the potential to reduce late-stage surprises, when problems discovered during verification can be significantly more expensive and time-consuming to resolve.
The process also supports evidence generation, which is particularly important in safety-critical industries. Engineering organizations must often demonstrate not only that software works but also how it was developed, tested and verified.
The prototype therefore aims to produce a traceable record of engineering activities and results.
Human Engineers Remain in Control
Despite the extensive use of AI agents, the project does not seek to remove engineers from the development process.
Human oversight remains a central principle of the architecture.
Critical decisions require engineering review and approval, while deterministic verification tools provide objective analysis. This is especially important in automotive applications where software can influence vehicle safety and reliability.
The approach recognizes that AI is most valuable when it augments engineering teams rather than replacing their responsibility.
AI agents can coordinate repetitive activities, analyze results and suggest improvements, potentially allowing engineers to spend more time on higher-level decisions and complex technical problems.
This model could become increasingly important as automotive software continues to grow in size and complexity.
Potential for Distributed Automotive Development
The prototype was also designed with potential deployment through AWS Marketplace in mind.
A cloud-delivered architecture could make advanced compliance and verification capabilities more accessible to distributed engineering organizations. Automotive development increasingly involves teams working across different geographic locations and organizational boundaries.
Cloud-based services could provide a common environment in which engineers, suppliers and technology partners can access relevant workflows and engineering capabilities.
For OEMs and major suppliers, this could potentially reduce the complexity associated with deploying specialized tools across multiple development organizations.
The hackathon prototype is therefore more than a demonstration of AI capabilities. It illustrates how cloud infrastructure, AI agents, semiconductor knowledge and established engineering tools could be brought together to create a connected development environment.
Highest Score for OEM Purchase Readiness
The project’s highest score for OEM purchase readiness is particularly notable.
Hackathons often focus on technical experimentation, but commercial readiness indicates whether a solution has a clear path toward solving real industry problems.
By receiving the highest score in this category, the project demonstrated that the judges viewed the workflow as having meaningful potential for practical adoption.
The prototype directly addresses several challenges faced by automotive manufacturers, including fragmented processes, increasing software complexity, verification requirements and the need for greater engineering efficiency.
Its emphasis on governance and human oversight also aligns with the requirements of industries where reliability and traceability are essential.
Broader Applications Beyond Automotive
Although the prototype was designed primarily around software-defined vehicles, the architectural principles demonstrated by the project could extend beyond automotive applications.
DLR’s participation highlights the potential relevance of the approach to aerospace and other safety-critical industries.
Aerospace and defense organizations similarly manage complex engineering processes involving strict requirements, extensive verification and certification evidence. Like automotive companies, they need to balance development speed with safety, reliability and regulatory compliance.
The same principles could potentially be applied wherever AI-assisted engineering must operate within controlled and auditable workflows.
These principles include trusted technical information, governed AI orchestration, deterministic verification, continuous testing, traceability and human approval.
AWS Recognizes Practical Industry Impact
AWS also highlighted the practical nature of the winning project.
Stefano Marzani, Worldwide Head of Emerging Technologies, Automotive and Manufacturing at AWS, said the winning team stood out by combining an understanding of software-defined vehicle challenges with the ability to bring multiple technologies together into an executable workflow.
That distinction is important as companies increasingly experiment with AI-powered engineering tools.
The automotive industry does not simply need AI systems that can generate code. It needs systems capable of operating within existing engineering, safety and compliance frameworks.
The TASKING-led prototype demonstrates one potential model for achieving this balance.
A Potential New Direction for Automotive Software Development
The success of the TASKING, Infineon and DLR team at the AWS hackathon illustrates how AI could play a more significant role in the future of safety-critical software development.
The project combines AI orchestration with established engineering practices rather than treating artificial intelligence as a standalone coding solution.
By connecting trusted semiconductor information with implementation, compliance, testing, coverage analysis and evidence generation, the prototype demonstrates a more integrated approach to the automotive V-cycle.
For vehicle manufacturers, such integration could eventually help reduce development bottlenecks, shorten feedback cycles and improve the consistency of verification activities.
At the same time, the emphasis on human oversight and deterministic verification recognizes that safety-critical development requires accountability and technical rigor.
As software-defined vehicles continue to evolve, automotive companies will need new ways to manage increasing software complexity without compromising safety or quality. AI-assisted engineering workflows could become an important part of that transformation.
The AWS hackathon win provides TASKING, Infineon and DLR with a strong platform for further exploring that opportunity. More importantly, it demonstrates a possible path toward an automotive development environment in which AI agents handle coordination and repetitive engineering activities while trusted data, deterministic tools and human engineers remain at the center of critical decisions.
The result is a vision of software development that is faster and more connected, but still governed by the engineering principles required for safety-critical systems.
Source Link:https://www.businesswire.com/
