SENIOR APPLIED AI ENGINEER / BERLIN

A practical open-source skill for long-running coding agents.

It keeps task_plan.md, findings.md, and progress.md on disk so work can continue after context loss, crashes, or compaction.

26K+ stars · 2.1K+ forks
18+ documented platforms

RUST / TAURI · TYPESCRIPT / BUN
NEO4J / GRAPHRAG

LinkedIn · Email · Writing

CLICK ANY VISUAL TO OPEN ITS SOURCE

• Snyk, Top 8 Claude Skills for Developers: lists planning-with-files first and marks its SKILL.md verified.
• OpenUI Lab: labels OpenUI Forge a Tool by Community.
• Daily Dose of Data Science: includes planning-with-files among 16 AI Agent Skills for AI engineers.
• Merged upstream work: 9 pull requests across 6 external repositories, including safe Windows reparse-point handling, agent observability, and German-language career tooling.

Work credited downstream and merged into external projects.

• Verified source credit · CCteam Creator: credits and links planning-with-files for its persistent three-file planning pattern.
• Windows filesystem safety · MemPalace #558: merged handling for unreachable Windows reparse points, so scans skip inaccessible entries instead of aborting.
• Agent observability · awesome-claude-skills #21: merged LangSmith trace retrieval for debugging LangChain and LangGraph agents.
• Skill-format compatibility · uberSKILLS #67: merged support for importing Claude Code skills without a required trigger field.

At migRaven in Berlin, I work on agent infrastructure and graph-backed identity, data, and access systems. These public products show the domain and systems context of that work:

• migRaven.MAX: identity, data, and access management with a Neo4j knowledge graph, enterprise connectors, and LLM-assisted querying
• aikux.Brain: enterprise knowledge graphs, GraphRAG, and semantic access to organizational knowledge
• migRaven.Archiver for Teams: Microsoft Teams archiving and visibility into external permissions

My role also includes internal workflow and script-orchestration systems within migRaven.MAX.

• Agent reliability: persistent state, context recovery, orchestration, audit trails, and release gates that stop on failed checks
• Local-first product engineering: Rust, Tauri, Svelte, Bun, browser extensions, privacy boundaries, and reproducible packaging
• Graph systems: Neo4j, GraphRAG, identity and access graphs, checkpoints, permissions, and enterprise retrieval

I also teach software development and AI engineering, including Python, agent-system design, and production engineering practices. That work has shaped how I document complex systems, explain tradeoffs, and collaborate across different levels of technical experience.

Technical depth, clear explanations, and communication across different experience levels.

Agent reliability, visual planning, music control, generative UI, model routing, and knowledge graphs.

Agent tooling, desktop applications, Rust systems, graph infrastructure, media tools, and web products.

Recognition, direct downstream credit, related implementations, merged upstream work, launch coverage, and teaching references.

Berlin · LinkedIn · Email · dev.to