Data Scientist

10/2023 – offen

As part of the development of new Pillar 2 (ICAAP/ILAAP) risk models, I was responsible for the end-to-end validation process. My role included

1) designing and documenting a robust validation framework,

2) implementing generalized validation analyses in Python to ensure reproducibility and scalability,

3) and compiling detailed validation reports.

The focus of my work was to critically assess model design, performance, and regulatory compliance, while establishing standardized procedures for future validations.

Confluence, Git, Jira, Python, SQL

3/2021 – 9/2023

Development of Probability of Default (PD), Loss Given Default (LGD), and Exposure at Default (EAD) models in accordance with the Internal Ratings-Based (IRB) approach. The project was implemented using Python and SAS and covered the following key areas:

1) Performing comprehensive data quality checks on the reference dataset to ensure consistency, completeness, and reliability for modeling purposes.

2) Creating robust modeling datasets for both estimation and calibration phases, ensuring the datasets were tailored to the specific requirements of each model component.

3) Applying and quantifying the Margin of Conservatism (MoC) in line with regulatory standards, incorporating appropriate adjustments to account for uncertainties in data and model design.

4) Conducting thorough model testing, including backtesting, validation, and sensitivity analysis, to ensure predictive performance and compliance with regulatory expectations.

The project emphasized a structured, transparent approach to model development, aiming to deliver models that are both technically sound and regulatory compliant.

Automatisierungstechnik (allg.), Datenmodelierung, Digitalisierung, Python, Risikomanagement, SAS (Software)

1/2021 – offen

In addition to my project work, I have independently built several Python-based applications:

1) Development of an AI agent (using Python) to automate manual email management, enabling user communication through a WhatsApp API. I created a webhook using FastAPI and deployed the solution live on AWS to seamlessly route user messages to the AI agent.

2) Creation of a web application for visualizing the Newton model, using Streamlit for intuitive and interactive user experiences.

3) Development of a risk model engine designed for deployment, offering scalable and efficient risk assessments.

4) Implementation of a web scraper to automate data extraction from various internet sources, supporting streamlined data collection processes.

Python