Description
As a Doctoral Researcher in Data-Driven Development Methods for Vehicle Diagnosis, you will be responsible for conducting independent literature research to analyze and evaluate the current state of technology. You will design data-driven methods by formulating hypotheses, defining evaluation criteria, and developing a roadmap. Your tasks will also include developing innovative solutions based on language models, such as Retrieval Augmented Generation (RAG), anomaly detection, and error cause analysis. Additionally, you will explore and implement prototype scalable mixture-of-experts models to improve anomaly detection in automotive control unit logs. You will build a prototype system for automated generation of explainable diagnosis rules using large language models and their inferential capabilities. Furthermore, you will develop data pipelines in the cloud that process diagnosis and protocol data (UDS, CAN, DoIP) securely and efficiently, including ingestion, quality assurance, and anonymization. You will validate your approaches using real data, create benchmarks, and provide evidence of effectiveness. You will also develop prototype tools and services for development, production, and after-sales, and document them comprehensively. Finally, you will create publications and ensure compliance with data protection, information security, and AI governance.