AI Engineering / Machine Learning Research / Data Platforms
Data & AI Engineer at dmTECH, where I build the data platform and agentic tooling. ML researcher at heart: my TU Berlin master's thesis combined graph neural networks and LLM-derived embeddings for ransomware detection. I like taking an idea from zero to something that works.
Trajectory
Production
At dmTECH I set up the department's Snowflake platform and build agentic Claude Code skills backed by cloud APIs, alongside the data pipelines and Terraform deployments around them on Google Cloud.
Research
My master's thesis explored how LLM-derived embeddings can enhance graph neural networks for Android ransomware detection, on a self-built dataset designed to avoid train and test leakage.
Teaching
I taught algorithms and programming at TU Berlin, and I run free programming workshops with Starcode Berlin to address the gender imbalance in computer science and STEM.
Selected work
Master Thesis
Master thesis on graph-based Android ransomware detection that enriches function call graphs with LLM-derived semantic node representations. Covers APK analysis, graph construction, semantic feature engineering from code artifacts, and GNN-based classification. Built a self-made Android malware dataset designed so that training and test data share as little code as possible, and used it to benchmark LLM judging, LLM summarisation, and code embedding models as sources of node features. Language models were self-hosted with vLLM.
Project Highlight
LangChain-based LLM agent to autonomously generate, compile, and debug Java plugins for Google's Tsunami Security Scanner. Designed a two-stage workflow combining code generation with compiler feedback for self-correction, enabling fully automated plugin development for OWASP Juice Shop vulnerabilities. In evaluation, the agent produced 16 working vulnerability detection plugins, all without manual code fixes.
Project Highlight
Geospatial Computer Vision pipeline for automated building detection from multispectral satellite imagery. Integrated Sentinel-2 and OpenStreetMap data, generated segmentation masks, prepared stratified training/validation datasets, and trained PyTorch-based CNN and U-Net models with hyperparameter tuning and augmentation.
Technical focus
I work most comfortably where machine learning, data engineering, and product-minded implementation meet. For a fuller picture, the CV page covers coursework, teaching, volunteering, and the broader toolset I worked with.
If the role needs someone comfortable moving between research context, implementation, and explanation, that is the space I like working in. I'm also open to early-stage collaborations.