Projects
A selection of projects showcasing my work in AI, Economics, and Data Science — from applied machine learning to published research.
Project in AI
Lung Disease Detection using X-Ray Images
-
Develop deep learning models (CNN and Transfer Learning) alongside classic machine learning models (XGBoost, SVM, Random Forest, Logistic Regression) to detect lung diseases from 21,000 chest X-ray images.
-
Optimize for high F1-score with a focus on recall, minimizing false negatives to enhance clinical reliability while balancing model interpretability and overfitting prevention.
-
Achieve 94% accuracy with the best Transfer Learning model using TensorFlow, reaching F1-scores of 96% for COVID and 97% for Viral Pneumonia, with a 93% recall rate for COVID.
-
Deploy an interactive Streamlit web app, enabling users to upload chest X-ray images for real-time predictions, supported by Grad-CAM visualizations for interpretability. Hosted on AWS Cloud for scalability.
Publication: Geographic diversification and bank lending during crises
Journal of Financial Economics, 2021 (with Dr. Sebastian Doerr)
​
We classify a large sample of banks according to the geographic diversification of their international syndicated loan portfolio. We show that diversified banks maintain higher loan supply during banking crises in borrower countries. Positive loan supply effects lead to higher firm investment and employment growth. Diversified banks are stabilizing due to their ability to raise additional funding during times of distress. Distinguishing banks by nationality reveals that diversified domestic banks are a stable source of funding, while foreign banks with little diversification are fickle. Findings suggest that declining financial integration makes countries more vulnerable to local financial shocks.
Publication: Makro-ökonomische Dynamik von Arbeitsmärkten
WSI Mitteilungen (with Dr. Eugen Spitznagel)
Die internationale Finanz- und Immobilienkrise hat in den USA ihren Ausgang genommen und eine weltweite Wirtschaftskrise mit tiefen Einbrüchen der gesamtwirtschaftlichen Produktion in den Industrieländern herbeigeführt. Die deutsche Volkswirtschaft war aufgrund ihrer ausgeprägten Abhängigkeit von den Exportmöglichkeiten besonders stark betroffen. Trotz eines tiefen Einbruchs der gesamtwirtschaftlichen Produktion erwies sich der Arbeitsmarkt in Deutschland als erstaunlich robust. In der relativ milden Rezession in den USA gingen dagegen viele Arbeitsplätze verloren, und die Arbeitslosigkeit stieg kräftig an.
Dissertation: The Real Effects of Banking Crises
This thesis examines the real economic effects of banking crises across three studies. The first shows that geographically diversified banks sustain higher lending during crises, supporting firm investment and employment, while foreign undiversified banks are the least stable lenders. The second finds that banks protect their core industries in crises but cut credit to others, with spillovers to non-crisis countries mitigated in main industries. The third reveals that government support, such as bank recapitalizations, increases home bias and distorts credit allocation toward larger, safer, and less innovative firms, while also expanding political influence over banks.
The Real Effects of Financial Protectionism
This paper analyzes the effects of government support of banks on European financial integration and firm outcomes using data on syndicated lending. Results show that banks increase their home bias in lending by 24.6%, after receiving government support. In turn, discriminated foreign firms can only imperfectly substitute this fall in lending by switching banks or issuing corporate bonds. Thus, the negative loan supply effect translates into lower sales and employment growth for foreign firms. In addition, government support distorts credit allocation in the home market by shifting lending to larger, safer and less innovative firms. Moreover, I document that politicians gain influence over banks by transferring control rights to the government as part of the support scheme. These results suggest that leaving bank bailouts at the national level within the European Banking Union, discourages international economic activity, distorts credit towards less productive firms and harms growth.