AI Engineer · Building local-first AI tools and the infrastructure to run them.
Based in Germany 🇩🇪 · Working toward relocating to Japan 🇯🇵
I build AI-powered applications end-to-end — from fine-tuning the models, to the serving layer, to the desktop UI on top. I'm especially interested in running capable models locally, self-hosted AI pipelines, and the messy infrastructure work that makes AI products actually usable.
I enjoy the parts of AI engineering most people would rather skip: Docker orchestration, GPU pipelines, service management, and keeping the whole thing running at 3 AM.
🌸 Hime — AI-Powered Light Novel Translation
A desktop application for translating Japanese light novels with a fine-tuned LLM pipeline.
- Stack: Tauri (Rust + TypeScript) · FastAPI · Python
- Models: Qwen family · Gemma family · LLM-JP · LFM2 — fine-tuned via Unsloth + LoRA
- Focus: Translation quality on literary Japanese, all models running locally
🚐 qamper-rent.de — Camper Rental Platform
Full web platform I designed and built for a camper rental service. Currently in maintenance mode.
Languages
AI / ML
Frameworks & Runtime
Infrastructure
- Fine-tuning translation models for Hime with Unsloth + LoRA on an RTX 5090
- Exploring local agentic workflows with Claude Code, MCP servers, and local LLMs
- Studying Japanese toward JLPT N2