• Several years producing technical documentation and maintaining knowledge bases (process docs, FAQs, onboarding material)
• Direct involvement in public-sector projects, delivering documentation and operator-facing content for federal/state programms
• Hands-on in IT-administration and -operation, planned and executed server migrations for over 600 employees at once, provided scripting support and runbook automation
• Practical experience with self-hosted infrastructure, especially Proxmox, Docker, Portainer, VPS hosting and troubleshooting containerised services
• Strong focus on implementable, ops-aware docs that reduce onboarding time and prevent production errors
• Currently working full time as a Technical Writer for multiple german government projects

• Professional degree as a Media Engineer/Technician
• Currently studying B.Sc. Computer Science (part-time)

I follow AI research closely and enjoy understanding models inside and out. My main interests are explainability, safe and privacy-aware deployment patterns, LLM behaviour analysis and practical AI applications.

• Member of the small and curated group of "Puzzlers" (community experts, product insiders and alpha/beta-testers) @ OpenAI
• Featured four times in the Community Spotlight @ OpenAI
• Deeply involved in prompt engineering, creating very advanced, in-depth prompts with techniques like SG-ICL, CoD, Cue-CoT, CoK and KNN
• Creating AI forensic tools to check for signs of AI generated content in texts, images, videos and audios
• Operating a local MoRE stack, running multiple specialized models in parallel on my pc to power automated multi-step troubleshooting including a generator, critic/verifier, summarizer, and retriever/embedder:
  • a MCP server exposes and governs tooling, retrieval and action endpoints
  • an orchestrator schedules requests, performs aggregation (vote/score), manages CPU/GPU offload and persists embeddings in a local vector DB (FAISS) to keep multiple experts low-latency and performant
  • leveraging a few Models I evaluate and deploy locally, mainly:
    • Mistral-7B as a high-throughput generator for draft generation and interactive response synthesis
    • Llama-3 (8B and quantized 13B variants) for instruction-following and medium-complexity reasoning / few-shot adaptation
    • Llama-4 Scout (quantized 17B active / MoE) for specialized long-context and multimodal verifier and deep reasoning expert
    • gpt-oss-20B for deep multi-step reasoning and as a critic/verifier for fact-checking and chain-of-thought verification
    • MPT-7B / Vicuna-7B for concise summarization, style adaptation and response refinement
    • sentence-transformers for dense embeddings and for semantic retrieval, similarity search and vector-DB indexing

• I've got two cats 🐱
• Living near Lake Constance (bordering Germany, Switzerland and Austria) on the German side
• 3D printing everything with my Bambulab P1S
• Love tinkering with both software and hardware, mostly:
  • self-hosted services
  • my homelab hosted on a mini PC
  • building web applications and AI forensic tools
  • running different LLMs or diffusion models
  • fixing electronic parts with precision mechanics tools and soldering station
  • mechanical keyboards

I try to learn something new every day.

• CPU: Intel Alder Lake N100 @ 3.40 GHz
• RAM: 16 GB DDR4 @ 3200 MT/s
• NVMe: 512 GB M.2 NVMe
• Mainboard: MW-N100

• Proxmox VE

• CPU: Intel Core i9-14900KF @ 6 GHz
• RAM: Crucial Pro 128GB DDR5 (2 x 64GB) @ 5600 MT/s
• CPU: NVIDIA RTX 4080 SUPER 16 GB
• NVMe: 6 TB Samsung 990 Pro M.2 NVMe (2 TB + 4 TB)
• Mainboard: MSI PRO B760-P WIFI

• Windows 11
• Arch Linux with Hyprland