Dr. Aiman El Asad

About me

I’m an industrial AI researcher with a background in physics, mathematics, software engineering and automotive production. Since 2022 I have been working at Mercedes-Benz AG in Sindelfingen, where I carried out my doctoral research within the SofDCar project in cooperation with the University of Stuttgart. My dissertation, AI-Based Software Update Release Impact Analysis, develops methods that estimate how a vehicle software update will affect production before it is rolled out. I passed my doctoral examination in July 2026 with the distinction magna cum laude.

My work sits at the intersection of three areas:

Software-defined vehicles — release documentation, commissioning and testing data, software traceability.
Production quality — First Time Quality (FTQ) as a measurable indicator of the impact of software updates.
Large language models — retrieval-augmented generation, agentic systems and semantic clustering of production error messages.

Before moving into industrial AI I studied physics at RWTH Aachen University (B.Sc. in experimental and M.Sc. in theoretical particle physics) alongside additional studies in mathematics, and I taught theoretical physics to bachelor’s and master’s students for three years. I still enjoy teaching: in spring 2026 I taught physics and mathematics for four months at a Gymnasium (German secondary school).

Outside work I like cooking, art, photography, plants and reading.

CV

A summary of my CV. A full version is available on request via the contact page.

Work experience

Doctoral Researcher & Research Scientist, Industrial AI and Automotive Software Mercedes-Benz AG, Sindelfingen, Germany Developed AI-based methods for assessing the production impact of automotive software updates. Designed data pipelines and evaluation methods for release documentation, commissioning data and production-quality indicators, and built LLM-based retrieval and agent systems. Applied research within the SofDCar project in cooperation with the University of Stuttgart and industrial partners.
Software Tester & Developer andagon GmbH, Cologne, Germany Developed and tested software components and automated test suites (aqua, Jira, STARC).

Education

Doctoral studies in engineering University of Stuttgart, Germany Dissertation: AI-Based Software Update Release Impact Analysis. Doctoral examination passed magna cum laude.
M.Sc. Physics RWTH Aachen University, Germany Specialisation in theoretical particle physics; concurrent studies in mathematics.
B.Sc. Physics RWTH Aachen University, Germany Specialisation in experimental particle physics.

Teaching

Physics and Mathematics Teacher (fixed-term) Christian-Rohlfs-Gymnasium, Hagen, Germany Independently planned and delivered eleven lessons per week (physics, plus mathematics support classes).
Tutor in Theoretical Physics RWTH Aachen University, Germany Taught bachelor’s and master’s students and led a semester-long tutorial in theoretical mechanics.

Skills

Programming: Python (NumPy, pandas, scikit-learn, PyTorch, TensorFlow), SQL, Spark, C++, CUDA, Java, LaTeX
Platforms & tools: Azure Databricks, Azure Machine Learning, MLflow, Git, CI/CD, MLOps, LangChain, LangGraph, Hugging Face, vector databases
AI & machine learning: Industrial AI, large language models, retrieval-augmented generation, agentic systems, explainable AI, NLP, clustering, data engineering
Automotive: Software updates, production quality, testing and commissioning, root-cause analysis, software traceability, release impact analysis, production data analytics
Certificates: ISTQB® Certified Tester · aqua ALM · English for Academic Purposes I (C1), University of Stuttgart
Languages: German (native), Arabic (native), English (C1)

Publications

Peer-reviewed publications, manuscripts under review and conference contributions. Where available, titles link to the publisher’s page (DOI).

First-author publications

Quantifying the Impact of Software Updates through First Time Quality in Automotive Production (opens in a new tab) A. El Asad, K. Köhler, M. Hahn, H.-C. Reuss SAE Technical Paper 2026-01-0786, Stuttgart International Symposium, 2026
Strategies for Improving Data Management in Automotive Testing and Commissioning Processes (opens in a new tab) A. El Asad, K. Köhler, M. Hahn, H.-C. Reuss SAE Technical Paper 2025-01-0277, Stuttgart International Symposium, 2025
Advancing Automotive Production: An LLM-based Impact Analysis for Software Updates (opens in a new tab) A. El Asad, M. Hahn, Y. Zhai, H.-C. Reuss Procedia CIRP, vol. 134, pp. 127–132, 2025

Under review

An Industrial AI Framework for Automotive Software Traceability S. Aghaei*, A. El Asad*, E. Daub, F. Ansari (*shared first authorship) Journal manuscript under review, 2026

Conference contributions

An LLM-Based Approach for Semantic Clustering of Vehicle Production Error Messages A. El Asad, M. Hahn, H.-C. Reuss, E. Daub Poster, Automotive meets Electronics & Control (AmEC) 2026, VDE/VDI-GMM, Dortmund

Co-authored publications

Production-Optimized Automotive Software Architecture: Theory and Applications (opens in a new tab) Y. Zhai, M. Beck, A. El Asad, K. Köhler, M. Hahn, E. Sax IEEE Transactions on Industrial Informatics, vol. 22, no. 5, pp. 3741–3752, 2026
Generating Machine-Processable Specifications from Natural Language Using Large Language Models in Automotive Commissioning and Testing (opens in a new tab) K. Köhler, A. El Asad, M. Hahn, H.-C. Reuss SAE Technical Paper 2026-01-0768, Stuttgart International Symposium, 2026

Projects

Open-source projects from my spare time. All code is on GitHub (opens in a new tab).

gpu-kernel-lab

Roofline-driven CUDA kernel optimisation. Four workloads — 2D stencil, reduction, softmax and SGEMM — are tuned in documented steps, each measured against the hardware ceiling it is bound by (SGEMM reaches 97% of cuBLAS on an RTX A2000 Laptop GPU). Hypotheses that turned out wrong stay in the code and are labelled as such — the method is the point, not the final numbers.

Source on GitHub

torch-roofline

What is your PyTorch model actually bound by? A per-operator roofline profiler: it runs a model once, intercepts every operator, records time, FLOPs and bytes moved, and places each op on the measured roofline of your device. Reports memory- vs compute-bound ops, fusion candidates and whether lower precision, torch.compile or bigger batches are the right lever. CUDA, Apple MPS and CPU.

Source on GitHub

Contact

The easiest way to reach me is by e-mail. I’m happy to talk about industrial AI, software-defined vehicles, LLM systems or teaching.

E-mail aimanelasad@yahoo.de
GitHub github.com/aimanelasad (opens in a new tab)

Legal notice

Imprint

This is a private, non-commercial website.

Dr. Aiman El Asad
Germany
E-mail: aimanelasad@yahoo.de

Privacy

Controller for data processing on this website is the person named in the imprint above.

Hosting. This website is hosted by GitHub, Inc., 88 Colin P. Kelly Jr. Street, San Francisco, CA 94107, USA (“GitHub Pages”). When you visit the site, GitHub’s servers automatically process technical data that your browser transmits (IP address, date and time of the request, requested URL, browser type and version, operating system, referrer) in server log files. This processing is necessary to deliver the website securely and is based on Art. 6 (1)(f) GDPR (legitimate interest in a secure and efficient provision of the site). GitHub is certified under the EU-US Data Privacy Framework. Details: GitHub Privacy Statement (opens in a new tab).

No tracking. This website does not use cookies, analytics, tracking pixels, embedded social media or content from third-party servers. The web font is served from this site itself.

Contact by e-mail. If you write to me, I process the data you provide (e-mail address, content of the message) solely to answer your request (Art. 6 (1)(b) and (f) GDPR). Messages are deleted once the matter is settled, unless legal retention periods apply.

Your rights. Under the GDPR you have the right to access, rectification, erasure, restriction of processing, data portability and objection (Art. 15–21 GDPR), and the right to lodge a complaint with a supervisory authority (Art. 77 GDPR).