Thesis work in Machine Learning 30 hp - Multimodal Modeling for Vehicle Fault Diagnosis

Introduktion:

Scania is undergoing a transformation from being a supplier of trucks, buses and engines to a supplier of complete and sustainable transport solutions. A critical success factor in the shift to becoming a world-leading provider of transport solutions is the operational availability of vehicles. The belief is that the importance of uptime will be even further accentuated going forward. 

Background: 

At Scania, fault diagnosis plays a central role in ensuring vehicle uptime, reliability, and customer satisfaction. Current diagnostic systems often rely on single-modality data (e.g. sensor signals or textual descriptions), which limits their ability to capture the full context of complex vehicle faults. Scania’s vehicles and service systems generate a rich set of multimodal data sources, including sensor measurements, fault codes, service records, images, and technician comments. The growing availability of multimodal data opens new opportunities to improve diagnostic coverage and model interpretability by learning from multiple perspectives simultaneously. 

Recent advances in multimodal machine learning, such as models like CLIP (Contrastive Language–Image Pretraining) and BLIP (Bootstrapped Language–Image Pretraining), have demonstrated the potential of combining visual and textual information for more robust understanding and prediction. Extending these ideas to the automotive domain could enable new ways of diagnosing complex faults by integrating text, image, and sensor data into a unified representation. 

Assignment:

This thesis project aims to develop knowledge and methods that advance data-driven diagnostics. By exploring how different types of data contribute to predictive performance, the project will help build more accurate, explainable, and intelligent diagnostic systems for Scania vehicles. 

The work will involve: 

• Investigating how diverse data modalities can be effectively fused and aligned. 

• Benchmarking state-of-the-art multimodal alignment and fusion models. 

• Evaluating model performance using established metrics and explainability analyses. 

• Building a multimodal pipeline that integrates different data types (text, image, sensor, etc.) for vehicle fault diagnosis. 

• Proposing potential improvements or extensions for future applications. 

Education And Skills:

We are looking for a master’s student in Machine Learning, Data Science, Computer Science, or a related field, who possesses knowledge or experience in: 

• Machine learning algorithms and deep learning concepts. 

• Python and machine learning frameworks such as TensorFlow or PyTorch. 

• Experience with cloud platforms such as AWS is considered a merit. 

An analytical, communicative, and collaborative mindset will ensure you thrive in our supportive and dynamic team. Fluency in English is required. 

Contact: 

Supervisor: Hsiao-Chun Hu, Advanced Analytics & AI Solutions (TGROSSA) 

hsiao-chun.hu@se.traton.com 

Application: 

Number of students: 1-2 

Start date: Spring 2026 

Estimated time needed: 20 weeks (30 credits), Full-time 

Language: English 

Location: Södertälje, hybrid work possible 

Your application must include a CV, personal letter and transcript of grades.  
 

A background check might be conducted for this position. We are conducting interviews continuously and may close the recruitment earlier than the date specified.