Christian Hobelsberger

Christian Hobelsberger

Statistics Student at LMU Munich | Data Science at Munich Re

Munich Re

relAI (M.Sc. Student)

Biography

I am a student at LMU Munich, majoring in Statistics & Data Science. I am currently working at Munich Re as part of an integrated study program and a M.Sc. student at the Konrad Zuse School of Excellence in Reliable AI. My interests include time series analysis, statistical models, and cost-effective machine learning algorithms.

In my free time, I dedicate myself to teaching chess to kids at my local school and building cool stuff.

Contact me to get a copy of my résumé.

Interests
  • Statistics
  • Machine Learning
  • Artificial Intelligence
  • Stock Market
  • Chess
Education

  • Master of Science (M.Sc.), Statistics and Data Science, 2026

    Ludwig-Maximilian-Universität, Munich

  • Exchange Semester, Master of Science (M.Sc.), Artificial Intelligence, 2024

    University of Groningen, Groningen, Netherlands

  • Bachelor of Science (B.Sc.), Statistics (Computer Science), 2023

    Ludwig-Maximilian-Universität, Munich

  • Abitur / GCE A-Levels, 2020

    Gymnasium Munich Moosach

Professional Experience

 
 
 
 
 
Munich Re
Integrated Study Program (Masters)
Munich Re
Sep 2024 – Present Munich
  • Underwriting Casualty Treaty (Financial Lines) (Oct 2025 - present)
  • AI Engineering (June 2025 - Sep 2025)
  • Underwriting F&C Casualty (Financial Lines) & Digital Underwriting (Mar 2025 - May 2025)
  • Project Manager for the Tech Trend Radar 2025 (Oct 2024 - Feb 2025)
 
 
 
 
 
Munich Re Specialty - North America
Modeling & Analytics Intern
Munich Re Specialty - North America
Jul 2024 – Aug 2024 New York, US
  • Developed an AI-powered system to automate the extraction of primary insurance submission information using state-of-the-art large language models (LLMs). Improved the F1 score for various variable extractions by up to 100%
  • Improved the performance of generative AI models by introducing advanced prompting techniques, leading to significant gains in the quality and efficiency of insurance information extraction
 
 
 
 
 
Munich Re
Central Analytics Intern
Munich Re
Sep 2023 – Jan 2024 Munich
  • Conception and implementation of a global three-part training series on generative AI and Large Language Models (LLMs) for more than 700 colleagues worldwide
  • Supporting the company-wide rollout of the low code/no code tool Dataiku, community management and leading the technical support of the business units for use cases, including automatic data cleansing and harmonization through LLMs and termination statistics
  • Participation in the Global Data & AI Conference 2023 with over 350 participants
 
 
 
 
 
Telefónica Germany
Data Science Working Student
Telefónica Germany
Jun 2021 – Sep 2023 Munich
  • Developed an application from scratch using Python and the Gitlab API, seamlessly enabling collaboration between teams for over 100 employees
    Tools used: Python, Gitlab-API, Docker, Gitlab-CI/CD, Scheduler
  • Retrained a machine learning model in Python for alarm generation on financially-relevant time series data, resulting in improved aggregation of alarm periods and a clear reduction of false alarms by 33%
    Tools used: Python (scikit-learn), SQL, Microsoft Azure Databricks
  • Compiled a large, labeled dataset from various data sources using complex SQL queries, and subsequently implemented automated plausibility checks using cloud computing resources
    Tools used: Python, SQL, Oracle/Apache Hive Database
 
 
 
 
 
GESIS - Leibniz Institute for the Social Sciences e.V.
Teaching Assistant
GESIS - Leibniz Institute for the Social Sciences e.V.
Aug 2023 – Aug 2023 Cologne
  • Assisted in teaching the course “Data Science Techniques for Survey Researchers” at Summer School
 
 
 
 
 
After-school care
Gymnasium Munich Moosach
Jun 2019 – Jun 2023 Munich
  • Professional support for students in grades 5-8 in various subjects
  • Management of a chess program that allows students to learn and explore the game of chess

Projects

*
All Data Science Mental Health Deep Learning Other
Systematic Evaluation of Uncertainty Estimation Methods in Large Language Models
This project investigates ways to improve the reliability of large language models (LLMs) by measuring their confidence in generated answers. Four uncertainty estimation methods were tested on question-answering tasks. These were VCE, MSP, Sample Consistency, and CoCoA. The results show that, although each method captures different aspects of confidence, the hybrid CoCoA method provides the most reliable and calibrated results overall.
github ArXiv
Systematic Evaluation of Uncertainty Estimation Methods in Large Language Models
Heart Disease Prediction and Interpretability
This project compares LIME and SHAP for explaining heart disease predictions using the UCI Heart Disease dataset. Several models were built from scratch, with XGBoost achieving the highest accuracy (83.3%). While SHAP produced stable, theory-based feature attributions, LIME offered intuitive, instance-level insights that benefited from an aggregated approach to improve consistency. Together, they highlight how combining interpretability tools can make machine learning in healthcare both accurate and transparent.
github
Heart Disease Prediction and Interpretability
Optimizing Blood Cell Classification with Convolutional Neural Networks
Biomedical image analysis plays a critical role in the diagnosis and treatment of patients with serious diseases. Advances in data availability and machine learning have improved the accuracy and efficiency of these analyses. This paper explores the use of convolutional neural networks (CNNs) for blood cell classification, a task essential for complete blood counts that identify infections based on blood cell characteristics. Manual assessment of blood cells is error-prone and resource-intensive, highlighting the need for automated solutions. We implemented a CNN using the BloodMNIST dataset, which contains over 17,000 images across eight blood cell classes. The CNN achieved superior performance with an accuracy of 0.914, significantly outperforming the baseline 1-layer Feedforward Neural Network on all evaluated metrics. Our results demonstrate that CNNs, with their ability to capture spatial hierarchies, provide a robust solution for blood cell classification, offering improved accuracy and reliability for medical diagnostics.
github
Optimizing Blood Cell Classification with Convolutional Neural Networks
Finetuning t5 transformer for Educational Question Generation
This paper presents a detailed study on the fine-tuning of Google’s t5 model aimed at enhancing its ability to generate high-quality, relevant questions. Leveraging the diverse and widely accepted SQuAD dataset, we trained our model specifically for question generation. Our results indicate significant improvements in multiple performance metrics, including BLEU, ROUGE and Word Mover’s Distance scores. The results were evaluated in a qualitative manner using an interactive interface. We discuss the implications of these findings for the field of model based question generation, highlighting potential applications in education. Despite the advancements, we acknowledge certain limitations in terms of generalization. Future work is suggested to further refine the model, explore alternative architectures, and integrate the enhanced model into broader AI systems.
Finetuning t5 transformer for Educational Question Generation
A Comparative Analysis of Prompting Techniques for Fallacy Detection
This study explores the effectiveness of different prompting techniques for fallacy detection, a task requiring complex reasoning, using three State-of-the-Art (SotA) LLMs: Llama3, Gemma 7B, and Mistral 7B. We evaluate five prompting techniques: a handcrafted baseline, Chain-of-Thought (CoT), Thread-of-Thought (ThoT), Knowledge Generation, and Self-Consistency, against the Multi-level Annotated Fallacy Dataset (MAFALDA). The study aims to determine the impact of these techniques on LLM performance in fallacy detection, identifying the best combination of LLM and prompting technique according to the MAFALDA taxonomy. The findings highlight the individual strengths and limitations of each prompting technique and discuss future research directions.
github
A Comparative Analysis of Prompting Techniques for Fallacy Detection
Multiple Choice Question Generation by Fine-Tuning BART in Modular and Unified Approaches
To streamline the generation of high-quality multiple-choice questions (MCQs), two sequential pipelines were created: one with a modular design for question/answer and distractor generation, and another with a single unified model. BART was fine-tuned on the SciQ dataset, and the unified model outperformed the modular model on qualitative measures and cosine similarity of correct answers and distractors. Another pipeline was developed to extend the use of the model: given a URL and a topic, it generates an MCQ. A fine-tuned BART-large model successfully generated MCQs beyond the science domain of the SciQ dataset.
github
Multiple Choice Question Generation by Fine-Tuning BART in Modular and Unified Approaches
Implementation and extended Replication: Unbiased Look at Dataset Bias
In our Dataset Bias Analysis project, we conducted a literature review and empirical study on dataset biases in image classification. Building on the work of Torralba and Efros in 2011, we reproduced their Histogram of Gradients (HOG) and Support Vector Machines (SVM) results and compared them with modern Convolutional Neural Networks (CNNs) using the caltech101, MSRC, and PASCAL VOC-2007 datasets. Our study focused on identifying selection, framing, and labeling biases, and we found that both SVM and CNN models showed significant performance drops when trained and tested on different datasets, highlighting the persistent problem of dataset bias affecting model generalization.
github
Implementation and extended Replication: Unbiased Look at Dataset Bias
The impact of pandemic restrictions on mental health during COVID 19 in Germany (Bachelor Thesis)
This project utilizes data from the Global COVID-19 Trends and Impact Survey (CTIS) and Corona data Germany to access the the impact of pandemic restrictions on mental health during COVID 19 in Germany.
PDF github
The impact of pandemic restrictions on mental health during COVID 19 in Germany (Bachelor Thesis)
Multivariate Analysis of Climatological and Hydrological Low Water Drivers in Bavaria (Statistical Internship)
In our statistical internship, we conducted multivariate analysis on climatological and hydrological low water drivers in Bavaria using simulated and measured data from three river gauges between 1990 and 2020, focusing on predicting low water events through Generalized Additive Models (GAMs) with driver-specific splines and variable-specific quantile distribution analysis, highlighting the importance of groundwater level, precipitation group, and soil water as drivers, observing distinct north-south differences, and exploring the relationships between extreme events and relevant variables.
PDF github
Multivariate Analysis of Climatological and Hydrological Low Water Drivers in Bavaria (Statistical Internship)
CTIS Mental Health
We use the UMD Global CTIS survey, policy response information as well as Google trends data provided from Our World in Data. As part of our exploratory data analysis we built a Shiny App displaying important mental health metrics from the data.
github
CTIS Mental Health

Accomplish­ments

Telefónica Germany GmbH & Co. OHG
See certificate
LinkedIn
See certificate
LinkedIn
See certificate

Skills

Microsoft 365

85%

English

80%

R

75%

Python

70%

SQL

70%

Java

55%

Contact