AI/ML Engineer · Physicist · Berlin 📍

I turn AI research into things that ship.
PhD in Physics · 3+ years building with LLMs · 9 hackathon wins across 2025–2026 · 110 public repos

⚡ See it running — no setup required

Live	What it is
hot-repos.netlify.app	GitHub landscape analyzer — surfaces trends across a starred corpus
github.com/kaiser-data/kitsune-mcp	MCP hub that shapeshift()s into 10,000+ servers at runtime
reaction-dynamics-berlin.netlify.app	Team-behavior evidence dashboard + interactive knowledge graph

�� About

role:        AI/ML Engineer
location:    Berlin, Germany 🇩🇪
focus:       LLMs · Agents · Automation · Document Processing · Edge AI
status:      Open to roles, projects, collaborations & hackathons
building:    110 public repos across agents, MCP, edge AI & automation
background:  PhD Physics → 3+ years shipping LLM systems

🏆 Hackathon Record

9 wins · 13 podiums · 2025–2026

2026

	Event	Track / Result
🥇	Cognee × Qdrant Hack Night Berlin	reaction-dynamics — solo entry
🥇	Agentic FinAI Hackathon Berlin	Dify Open Track
🏆	AI Agents Hackathon Berlin	Blockchain for Good Alliance Bonus Track
🥈	AI Builders Open Claw Hackathon	Open Claw
🥉	Needle Hackathon	Main Track
4️⃣	Superchat × Needle AI Builder Event	Salesbot
🏅	Cognee AI-Memory Hackathon	Top 7 Finalist

2025 — 6 wins · 8 podiums _{(click to expand)}

	Event	Track / Result
🥇	ElevenLabs Hackathon	n8n Category
🥇	NEAR AI Hackathon	Shade AI-Agent
🥇	AI Hardcore Hackathon	LLM Prediction
🥇	Masterschool Hackathon	SMS-Service
🥈	Distributed AI Hack Berlin	Open Track
🥈	IronHack DevOps Hackashow	DevOps

🎯 What I Build

Generative AI & LLM applications — agents, RAG, document pipelines
Agent & MCP tooling — MCP hubs, A2A messaging, observability, memory stacks
Automation workflows — n8n, LangGraph, Dify, custom orchestration
Edge & self-hosted AI — local-first inference across a private device mesh
End-to-end products — from idea → prototype → production

🗂️ Selected Projects

_{A slice of 110 repos — grouped to show the range.}

🧩 Agents & LLM apps

• reaction-dynamics — 🥇 team-behavior analytics from live Slack reaction timing; KS-test classifier + Cognee/Qdrant knowledge graph
• AngebotsBot — AI deal scanner: kaufda.de + Qwen VL 32B + Supabase + LangGraph + Chainlit + Telegram
• fraudmind — deterministic fraud-hunting over GDPdU audit dossiers; 22 rule families + Cognee knowledge graph, "no number without a source"
• carlover — multi-agent automotive assistant (LangGraph + ADAC data + Featherless AI)
• multi-agent-research — research automation with LangGraph + Claude + SerpAPI

⚡ Edge & self-hosted AI

• jetson-headless-inference — Jetson Orin 8GB headless AI API with model switching & power management
• comm-buddy (private) — local-first meeting analysis: whisperX + pyannote + Ollama across Mac + Jetson over Tailscale
• mac-ram-manager · cleandrive — zero-dependency macOS system tools

📊 ML & benchmarking · 🔁 Automation & infra _{(click to expand)}

📊 ML & benchmarking

• featherless-bench — systematic LLM benchmarking + DeepEval; Pareto frontier, Vision Arena, 19 text models
• llm-finetune-kit — beginner-friendly LLM fine-tuning: 3-line training, web UI, smart defaults
• berlin25-eurosat — federated learning (Flower) on EuroSAT satellite imagery

🔁 Automation & infra

• inbox-inferno-n8n — AI email agent with LLM-as-judge — scored 20/20
• free-llm-proxy-router — Go OpenAI-compatible proxy routing across free-tier providers (Groq, Gemini, OpenRouter, Cerebras…)
• aws-terraform-devops-infrastructure — production-grade AWS VPC architecture with Terraform IaC
• kaiser-echo — bilingual voice agent with memory (React + Cloudflare Workers)

🤖 How I Work with Claude Code

Claude Code is my daily driver. I run a Pro subscription and squeeze roughly 30× the usual mileage out of it by spending tokens deliberately — pairing the agent with memory frameworks like claude-mem and modern agent harnesses like ECC and superpowers to reuse context, cut redundant re-reads, and keep long sessions cheap. Effective spend over raw spend.

I instrument my own sessions with claude-code-langfuse-tracing, so these are real numbers from the last 30 days:

Traces	Observations	Tokens (30d)	Models in rotation
1.45K	18.4K	~1.7B	Opus 4.8 · Sonnet 5 · Haiku 4.5 · Fable 5

Things I've shipped for the Claude Code ecosystem:

• 🦊 kitsune-mcp — shape-shifting MCP hub: shapeshift() into 10,000+ MCP servers at runtime, one entry point, no restarts
• 🔭 claude-code-langfuse-tracing — zero-intrusion observability; a Stop hook replays transcripts into Langfuse as backdated, session-grouped traces with token/cost
• 📡 claude-A2A-Comm — A2A-protocol messaging between Claude Code sessions, local or over a Pi/VPS hub
• 🧰 marty-skills · kogitsune — personal skills, commands & harness for Claude Code

💡 Tech Stack

Core

AI & GenAI

Data & Infra

Edge

🎭 Beyond Code

🎤 VP Public Relations · Toastmasters Club Berliner Rhetorikmeister
♟️ Teams Organizer · Chessclub Weisse Dame e.V. (200 members)

📊 GitHub

_{💬 Berlin-based · open to AI/ML engineering roles, contract work and hackathon teams — martinkaiser.bln@gmail.com}