At 1648 Factory GmbH, I architect and deploy asynchronous LLM agents for procurement tasks, including tender analysis, document review, and Bill of Materials generation. The systems cut manual processing by an estimated 30–40%.
I also design Hybrid RAG and GraphRAG systems for technical document retrieval and product recommendations. I build ingestion pipelines and automated evaluations to assess retrieval quality, grounded answers, and agent task success.
At Genesis Tech, I owned ranking and recommendation systems for more than 1M users, achieving a 25% CTR uplift in A/B tests with under 50 ms inference latency. I also developed and deployed fraud detection, forecasting, and other machine learning models.
At Kernel Holding, I optimized transportation planning with linear programming, reducing logistics costs by 15%. My research project compared transfer-learning approaches for plant image classification, with EfficientNet reaching a macro F1 score of 0.92.

