Freiburg, Baden-Württemberg, Germany
Data Scientist in the Data & Analytics Services team at Haufe Group, focusing on Machine Learning, automation and cloud-based data platforms. - Designing and deploying Data Science and Machine Learning solutions in modern cloud architectures (Snowflake, Tableau) - Developing and evolving our automation, NLP and AI/MLOps technology stack - Building scalable data pipelines and ML systems for production use - Contributing to cloud migration initiatives and modern data platform architecture - Applying statistical methods, experimentation and machine learning to create measurable business impact - Collaborating with cross-functional data and product teams to deliver innovative analytics solutions
- Supporting the Central IT Office in data-driven projects and initiatives - Building and maintaining data pipelines for forecasting and analytics - Gaining practical experience in data science, including data preprocessing, modeling, and evaluation - Contributing to various projects that improve data availability and support business decision-making
Worked at the intersection of unsupervised deep learning and astrophysical spectroscopy to improve the characterization of stellar activity in high-resolution spectra from AU Microscopii (AU Mic), a young and active M-dwarf star. Main focus: Designed and implemented a modular pipeline using Variational Autoencoders (VAEs) to compress stellar spectra and uncover low-dimensional representations of stellar variability (e.g., starspots, faculae). The goal was to disentangle activity-induced radial velocity (RV) noise from potential exoplanet signals. Key contributions: - Preprocessed telluric-corrected CARMENES VIS_A spectra: RV-shift correction, spectral cropping, normalization, and interpolation. - Developed several 1D Convolutional VAE architectures (incl. residual and U-Net-based designs) in TensorFlow/Keras. - Regularized latent space via KL-annealing, dropout, and SNR decorrelation techniques. - Visualized and evaluated latent structure using UMAP, Lomb-Scargle periodograms, and phase folding. - Demonstrated structured ~90-day periodicity in the latent space, consistent with stellar rotation. - Achieved denoised reconstructions preserving spectral line integrity, enabling further astrophysical analysis. Tools and Technologies: Python, TensorFlow, Keras, Astropy, UMAP, SciPy, Spectral preprocessing, Poetry, GPU compute, PyTorch, NumPy, Matplotlib
I lead conversation hours for advanced German students (spring 2024) and taught two labs for beginner-level German students (winter and spring 2024), each with 8 to 12 students, twice a week. Additionally, I taught two labs for intermediate-level German students (fall 2023), also twice a week. I was responsible for organizing weekly events to help German students practice their speaking skills and preparing events related to studying abroad. (Department of German Studies - Prof. Sederberg, Ph.D., Prof. Bryant, Ph.D., and Prof. Powers, Ph.D.)
I conducted tutorials in Microeconomics 2 (Bachelor, summer 2023), both online and on-site, as well as tutorials in Microeconomics 1 (Bachelor, winter 2022/23). Additionally, I managed administrative tasks, including overseeing the Moodle course and responding to student emails. Collaborating with doctoral students was a valuable experience. (Chair for Markets, Competition & Institutions - Prof. Dr. Bruttel)