Applied AI engineer · researcher
I build machine-learning systems and study how to make them useful, reliable, and understandable.
I currently work on agentic AI and document intelligence for finance and insurance. My day-to-day sits between research and engineering: reading papers, designing evaluations, building production systems, and finding out where models fail.
Before generative AI, I worked across NLP, speech, computer vision, and adversarial machine learning. I studied Computational Linguistics at the University of Stuttgart and have been building with ML since 2018.
A few projects that reflect how I work across applied AI, research, and open source—from production document systems to computer vision and speech.
I build systems that read complex financial and insurance documents, retrieve the right context, coordinate specialized agents, and return structured outputs that can be evaluated and audited. My work covers OCR, RAG, orchestration, validation, monitoring, and production APIs.
GlyphNet detects phishing domains that look visually similar to trusted websites. It renders domain names as images and uses an attention-based convolutional neural network to identify homoglyph attacks that text-only methods can miss.
SpeakerDiff explores privacy-preserving speech representations with diffusion models. It generates speaker embeddings designed to retain useful speech characteristics while reducing identifiable speaker information.
Public datasets I have created for financial language modelling and visual cybersecurity research.
Notes on ideas I am learning, systems I am building, and questions that remain unresolved.
Agentic AI, document intelligence, and decision systems for finance and insurance.
Model robustness, adversarial ML, security, fairness, bias, and toxicity evaluation.
Temporal and geographic representations for language understanding.
NLP, machine translation, biomedical NER, speech, diarization, and multimodal learning.
I play table tennis and chess, cycle around Stuttgart, follow financial markets, and read about quantum physics. I am usually carrying a notebook—digital or otherwise.