EY Tax GmbH Steuerberatungsgesellschaft, Berlin

EY Tax GmbH Steuerberatungsgesellschaft, Berlin

• Led the design and development of a Next.JS web application for processing, querying, and visualizing business transactions, enhancing data accessibility.
• Developed and maintained a tax determination engine in C# and ASP.NET, improving data quality analysis, classification, and tax compliance for SAP business transactions.
• Integrated an AI copilot chatbot (Azure OpenAI, Langchain): enabled natural language to MongoDB aggregation, automated tax determination, VAT-ID validation.
• Conceptualized and co-engineered AI-driven agents for extracting, mapping, and analyzing structured and unstructured data. These agents facilitated seamless data integration across multiple sources, empowering our clients with faster insights and more informed decision-making.
• Reviewed code, performed pull request reviews, and led design discussions to ensure maintainability and performance.
• Mentored and managed a team of 3 junior consultants, fostering technical growth and team efficiency.

Ernst & Young GmbH Wirtschaftsprüfungsgesellschaft, Berlin

• Optimized a backend server (JavaScript, GraphQL, MongoDB) for real-time transaction queries.
• Co-developed a Vue.js web dashboard to monitor, filter and review SAP business transactions.
• Migrated a legacy desktop application to scalable ASP.NET web architecture, improving data transfer, analysis pipelines, and third-party API integrations.
• Strengthened quality assurance with 80+ unit tests (XUnit) and rigorous code reviews.

Ernst & Young GmbH Wirtschaftsprüfungsgesellschaft, Berlin

Technische Universität Berlin (Grade 1.2)

• Master thesis: "Multi-domain Sentiment Analysis with an Active Learning Mechanism". The thesis proposes three model architectures for binary sentiment analysis on multi-domain textual data using an Active Learning mechanism.

Technische Universität Berlin (Grade 1.5)

• Bachelor thesis: "An Analysis of Lightweight Cryptographic Algorithms".

Katerina Katsarou, Nabil Douss, and Kostas Stefanidis. 2023.

• REFORMIST: Hierarchical Attention Networks for Multi-Domain Sentiment Classification with Active Learning. In Proceedings of ACM SAC Conference (SAC’23). ACM, New York, NY, USA, Article 4, 10 pages.