Schaeffler and Raiqon present at ReConf: AI from POC to Production

When Tobias Sommerfeld (Schaeffler) and Szabi Koppany (Raiqon) took the stage at REConf 2025, the room was packed. With over 110 attendees adding the session to their agenda and an impressive average rating of 4.5 out of 5 stars, their presentation clearly struck a chord. Representing Raiqon in connection with the Schaeffler-supported research initiative KIMBA, the duo brought critical insights into the future of AI in requirements engineering and model-based systems engineering.

If you work in systems or requirements engineering, this session offered a rare combination of technical depth and real-world relevance. What made this talk stand out was not just the data, but the candid look at what it takes to move AI from research and prototypes into the heart of automotive development. Keep reading to discover the key insights and why this presentation received such a strong response from industry professionals.

Get the Slide Deck

High Engagement and Strong Feedback

REConf, one of the most important conferences for Requirements Engineering in the German-speaking world, saw outstanding interest in this joint presentation. The feedback highlighted both the relevance of the content and the clarity of the speakers:

Tackling Industry Pain Points with AI

The presentation opened with a stark reality: the automotive industry is facing intense pressure. Growing complexity due to software-defined vehicles, regulatory burdens, and a shrinking talent pool are all pushing development teams to their limits. As the speakers noted, “Nearly all companies see the need to improve their requirement quality.”

An example highlighted a recall involving over half a million vehicles by a major German OEM (the name was withheld to protect the innocent). Inconsistent requirements regarding the Isofix attachments led to this problem, with resulted in costs exceeding 100 million euros. Clearly, the stakes are high.

The KIMBA Project: A Strategic Alliance

The foundation of the talk was the KIMBA research project. The acronym stands for “Künstliche Intelligenz für System-Model-Bildung und Anforderungs-Management” (AI for systems modeling and requirements management). In this project, Raiqon collaborates alongside key industry players, including BMW Group as project coordinator. Schaeffler plays a central role in this initiative by shaping practical use cases and validating AI-based approaches in real-world scenarios.

KIMBA targets three main areas:

1. Automated model generation from textual requirements
2. Interpretation and comparison of documents for compliance and consistency
3. Evaluation of language models in real-world automotive contexts

The project is funded until 2027 and is set to redefine how AI supports requirements and system modeling.

Key Insights: AI for Requirements Engineering

Raiqon and Schaeffler’s contribution to KIMBA focuses on practical AI applications in product development. From identifying requirement inconsistencies to extracting formal documentation from raw code, their approach is deeply rooted in real use cases.

A standout case study demonstrated how 800 structured requirements were generated from code in just three hours—delivering over 80% in time and cost savings. This AI-supported workflow helps engineers focus on actual development instead of time-consuming documentation tasks.

Small but Mighty: The Advantage of Smaller Language Models

The presentation also addressed the evaluation of different language models. If well-trained, smaller models outperformed their larger, more expensive counterparts in key metrics like precision and recall. This is especially relevant for automotive applications, where determinism and explainability are critical.

“Smaller models are not just cheaper to run—they’re often better suited for domain-specific tasks.”

From POC to Production

A recurring theme in the presentation was the difficulty of moving from proof-of-concept (POC) to production-ready systems. Many AI projects stall due to a lack of scalability, integration hurdles, or compliance concerns. Raiqon emphasized its focus on secure, scalable, and cost-efficient deployment options—on-premise, private cloud, and integration with tools like Codebeamer.

The goal is to embed AI where it adds value without disrupting established workflows.

Building Bridges: Startup, Industry, and Academia Collaboration

One of the most impactful undercurrents of the presentation was the unique collaboration model behind the KIMBA project. By bringing together startups like Raiqon, established industrial players like Schaeffler, and leading research institutions, the initiative exemplifies a new way forward in tackling complex technological challenges.

Startups contribute agility, speed, and innovative thinking. Industry players contribute the scalability, real-world use cases, and infrastructure needed to validate AI solutions at scale. Academia brings methodological rigor and a wealth of research insights. In KIMBA, this triangle of collaboration proves to be greater than the sum of its parts.

“We are not building AI tools in isolation—we are co-creating them with the people who will use and govern them.”

Szabi Koppany, CEO Raiqon

This cooperative model ensures that the AI solutions being developed are not only cutting-edge, but also practical, robust, and aligned with both business needs and regulatory frameworks. It is a textbook example of how cross-sector collaboration can accelerate the path from proof-of-concept to production.

Looking Ahead

The collaboration between Raiqon and Schaeffler through the KIMBA project continues to push the boundaries of applied AI in engineering. As the speakers concluded, success in this field will depend not on chasing trends, but on solving real, high-value problems with sustainable, explainable technology.

And if the REConf 2025 response is any indicator, this approach is gaining the recognition it deserves.