Doctoral Research Associate in Generative Multimodal Recommender Systems
15.09., Wissenschaftliches Personal
The Chair of Software Engineering & AI is searching for a new doctoral research associate (TVL-E13) in in Generative Multimodal Recommender Systems
Research Focus
Recommender systems are a cornerstone of modern digital platforms, but traditional approaches like matrix factorization struggle with significant limitations, including the cold-start problem, static user representations, and data sparsity. While deep learning models offer improvements, they often come with high computational costs and require frequent retraining, which limits their scalability and adaptability.
This research project aims to pioneer the next generation of recommender systems by moving beyond static methods. The core of this PhD is to develop a generative multimodal recommender system that leverages pretrained multimodal encoders (like CLIP) and advanced sequential modeling techniques. The central hypothesis is that a user's preferences are encoded in the sequence of items they interact with. By representing items through rich multimodal embeddings (from images and text) and modeling user behavior as a sequence, the system can dynamically adapt to new content and users without constant retraining. This project will be conducted in close collaboration with our industrial partner, Audi, focusing on real-world application scenarios, such as recommending Points of Interest (POIs).
Your Tasks
Your work will focus on addressing key challenges at the intersection of recommender systems, generative AI, and multimodal representation learning. Your primary research questions will include:
1. How can we effectively model user preferences by analyzing sequences of multimodal item representations?
2. What are the most effective sequential models (e.g., Transformers, GRUs) for capturing temporal dynamics and generating accurate user behavior embeddings from a sequence of items?
3. How do different training strategies—such as sequential prediction, contrastive pretraining (e.g., InfoNCE loss), and joint optimization—impact the model's accuracy, robustness, and generalization capabilities?
4. Can fine-tuning a pretrained model like CLIP on domain-specific data (e.g., automotive or travel-related POIs) significantly enhance the quality of item embeddings and subsequent recommendations?
5. How can we design and implement a robust pipeline for data engineering, including the acquisition, cleaning, and preprocessing of multimodal data for an industrial use case?
6. What are the most suitable metrics (e.g., NDCG, hit rate) for validating the proposed system, and how does it perform when benchmarked against traditional and deep learning-based baselines?
In collaboration with Prof. Chunyang Chen and Dr. Marcel Christopher Gauglitz, Dr. Florian Meyer (Audi), the successful candidate will develop and apply novel methods in AI and machine learning, contributing directly to state-of-the-art research with high industrial relevance.
Your Qualifications
7. A strong background and Master's degree in Computer Science, AI, Machine Learning, or a related field.
8. Excellent programming skills in Python and experience with ML frameworks (e.g., PyTorch, TensorFlow).
9. A solid understanding of recommender systems, deep learning, and sequence models (e.g., Transformers).
10. Experience with or strong interest in multimodal machine learning (e.g., CLIP), reinforcement learning is highly advantageous.
11. A passion for tackling challenging research problems and developing innovative solutions.
12. Experience in interdisciplinary research within a collaborative, international environment.
13. Excellent command of English (written and spoken); the research and teaching language is English. German language skills are a plus for industrial collaboration.
14. Willingness to contribute to university teaching activities.
We Offer You
At TUM Campus Heilbronn, you'll find an exciting and challenging project within a dynamic and collaborative research environment. Our long-standing and internationally recognized expertise in the field, coupled with close collaborations with academic and industrial laboratories worldwide, ensures access to rich, high-quality experimental data based on state-of-the-art technologies. On top of this scientific environment, the candidate will also be part of the Audi AG as an external employer. The successful candidate will be integrated in the “AI Hub and data analytics team” of the Audi AG and work closely with technical experts to improve the future product.
Salary is paid according to remuneration group E13 TV-L (%) of the pay scale for the German public sector. This is a position for a Doctoral student. TUM is an equal-opportunity employer. Therefore, women are especially encouraged to apply. We hope this position can start in late .
Applications
Complete applications should be sent to. To ensure your application is directed correctly, please use the following subject line format for your email: [Doctoral Application] Generative Multimodal Recommender Systems. Please include a CV, a cover letter explaining why you are interested in the position and how you fit the profile, a brief summary of previous work experience, and the contact information of at least two referees.
The due date will be Oct 3,, and you may still submit the application after the due date until the position is filled in.
The position is suitable for disabled persons. Disabled applicants will be given preference in case of generally equivalent suitability, aptitude and professional performance.
Data Protection Information:
When you apply for a position with the Technical University of Munich (TUM), you are submitting personal information. With regard to personal information, please take note of the Datenschutzhinweise gemäß Art. 13 Datenschutz-Grundverordnung (DSGVO) zur Erhebung und Verarbeitung von personenbezogenen Daten im Rahmen Ihrer Bewerbung. (data protection information on collecting and processing personal data contained in your application in accordance with Art. 13 of the General Data Protection Regulation (GDPR)). By submitting your application, you confirm that you have acknowledged the above data protection information of TUM.