Senior AI Engineer · 20+ years building software · Kraków, Poland 🇵🇱
I build production LLM tooling — internal AI agents, RAG systems, multi-agent workflows, and AI-powered automation. Two decades of engineering across Python, TypeScript, and Rust gives me a different angle on AI work: I care about what's reliable, debuggable, and actually useful — not just what's clever.
🟢 Open to remote AI Engineer roles — and always up for a small project or collaboration. 🌐 More at lipinski.work · 💼 LinkedIn
brain.md — local-first markdown notes with a first-class MCP server. 16 tools + 2 resources for Claude Code, Claude Desktop, Cursor, and any MCP agent. Per-folder permissions, semantic RAG via LanceDB, single Bun binary.
curl -fsSL https://raw.githubusercontent.com/mi4uu/brain.md/main/install.sh | bashYour notes stay yours — on your machine, in plain markdown, readable and writable by AI agents you control. AGPL-3.0.
Real-world use case: I run a personal agent ("Hermes") wired to brain.md — spun up from WhatsApp, Gmail, a calendar, and local models via Ollama. It watches my channels, filters the noise, and surfaces only what matters — running my day for the cost of electricity. brain.md is the memory layer underneath it.
AI / LLM — RAG · multi-agent workflows · MCP · prompt engineering · eval pipelines · LangChain · Ollama · vLLM · llama.cpp · OpenAI / Anthropic APIs · local model deployment
Rust — Axum · SQLx · Tokio · Utoipa · Leptos · Candle
Python — FastAPI · Pydantic · SQLAlchemy · Pandas · data pipelines
TypeScript — Node.js · Bun · NestJS · React
Data & infra — PostgreSQL · DuckDB · Redis · LanceDB · Docker · Kubernetes · AWS
Auth & security — OAuth2 / OIDC · RBAC · SSO · secure third-party integrations
🧠 brain.md — MCP-native notes app · RAG · LanceDB · single Bun binary
🤖 botmarley-book — automated trading engine with a visual strategy editor and a high-performance DuckDB-powered backtesting engine; exchange-agnostic execution (Rust · Axum · DuckDB)
📝 webOctoScribe — browser annotation tool for capturing and exporting web sessions to train AI (Python)
🔭 appbeholder — lightweight, OTel-native app observability for when you don't want to stand up a whole Prometheus stack (Rust)
🦀 Bun (oven-sh/bun) — long-term contributor; currently refactoring AI-generated unsafe blocks into idiomatic safe Rust during the Zig→Rust rewrite
- 🌐 lipinski.work — portfolio + full CV
- 💼 linkedin.com/in/michallipinski
- 📍 Kraków, Poland · English (C1) · Polish (native)
Looking to build AI-powered internal tools, or want an engineer who knows where LLMs work and where they fail? Let's talk.