Course: Natural Language Processing (NLP)

• Responsibilites: Module Leader, Module Lecturer, Exam-Setting for a class of 25 MSc students

Curriculum (15h):

• Introduction to NLP and preprocessing pipelines
• Vectorization methods and machine Learning for NLP
• Word Embeddings and Text Similarity
• Sequence-to-Sequence Deep Learning models

Master thesis title: “Time Series Based Anomaly Detection For Fleet Connectivity”

Grade: 20/20 (100%)

• Development and evaluation of statistical and ML-based methods to detect vehicle connectivity anomalies in the networked ConnectedDrive fleet.
• Development of ETL pipelines in the context of Big Data to provide data for analysis.
• Design and development of dashboards for visualization and stakeholder-adequate reports.
• Presentation of the evaluated methods and results in project and management rounds.

Building an Aspect-Based Sentiment Analysis pipeline

• Designed and implemented an NL¸ pipeline combining topic extraction and sentiment analysis of customer reviews
• Applied the pipeline to analyze and gain insights from customer feedback in French and German.

Bachelor Thesis Title: “Multilingual Text Classification using Transformers”

Grade: 17.67/20 (88.4%)

• Fine-tuned an XLM-RoBERTa (Goyal, 2020) model using Transfer Learning techniques on 110.000 documents
• Implemented Temperature Scaling (Guo, 2017) to produce calibrated probability outputs, enabling uncertainty estimation
• Achieved an F1-score of 93% across 14 classes in 3 languages (German, English and French)
• Deployed an active learning web application to facilitate collaborative data labeling and efficient model training