At the German Aerospace Center (DLR), I lead engineering for agentic AI platforms, from architecture through delivery. One platform is in production use in a policy advisory project for a German federal ministry; I built modular multi-agent pipelines and a model-agnostic LLM gateway, with grounded citations and human approval gates for high-stakes outputs.
I also shipped a multilingual voice-to-image app using Whisper, SDXL, FLUX, and ControlNet, with fine-tuned LoRA adapters and quantized inference. For my M.Sc. thesis, I combined LLM spatial reasoning with a modified MeshGPT transformer for text-to-3D scene generation, cutting inference latency by 29–39% and achieving 99% semantic layout precision.

