Lester Hrabovszky, AI & Automation Engineer

01 / About

“Sometimes you gotta run before you can walk.”

The age of knowledge is over. Everything is searchable, everything is accessible. What counts now is what you do with it: how fast you move, how well you adapt, and whether you're asking the right questions before you start building.

That's the lens I bring to every project. AI agents, workflow automations with tools like n8n or Claude routines, Docker infrastructure. Built to work, built to scale, built to last.

02 / Systems

Container-Operator Fleet

Autonomous AI operators on one home server
Identical container blueprint: only persona, credentials and workflows differ
Three live: shipping code, outbound content, opportunity scouting
Ephemeral runs, nothing sits idle holding state

3 operatorsidentical container blueprint6 orchestration workflows eachidle cycles cost nothing

lancelot

Four-Layer Memory

Recent turns, long-term memory, tonal reference, stable facts
Character is ingested from text and lands as nuance in every prompt via vector retrieval
Overnight dreaming pass consolidates short-term memory into long-term memory

4 memory layersnightly consolidationone embedding contract across all agents

Second Brain

Notion knowledge system on a PARA structure
Custom Claude Code skills: capture, retrieval, duplicate checks
Automated daily research
Built for myself first, colleagues at work run on it now

4 linked databasescustom capture and query skillsdaily automated research

Knowledge Graph

Every agent memory, note and reference segment in one graph database
Vector search on top: one store, many consumers, separated by property
Operators know things instead of looking them up

graph and vector in one storeshared across the whole fleet

Currently building a production pipeline where an AI operator handles research, script and voice, and every video goes through a human review loop before publishing.

03 / Selected Work

Lesterhr/ai-promptbook-v2

AI Promptbook v2

Desktop app for managing AI agent instructions, system prompts and project templates in one place.

Prompts and agent instructions were scattered across repos, notes and chat history. This keeps them versioned and reusable.

ReactTypeScriptTauriRust

qro.lesterhr.com

Q-RO

Voice dictation that turns speech into clean, punctuated text anywhere on the system. Local transcription, AI cleanup, three recording modes on configurable hotkeys.

Typing was the bottleneck on capture. Most of what ends up in my knowledge system now starts as speech.

TauriReactPythonfaster-whisperClaude

Lesterhr/TaskWidget-for-Notion

TaskWidget for Notion

An always-on-top desktop widget that shows your Notion tasks as a floating sticky note.

Tasks that live behind a browser tab do not get done.

ReactTauriNotion API

04 / Capabilities

AI / Machine Learning

Models, agents and applied intelligence

Skill & Agent Building

Custom Claude skills, routines and agent systems

n8n Workflow Automation

Self-hosted automation pipelines

Docker & Self-Hosting

Containerized services on my own hardware

Second Brain in Neo4J or Notion

Connected second-brain systems

AI Assisted Coding

Shipping faster with AI pair programming

Custom Software

Desktop and web tools built to purpose

System Architecture

Built to work, built to scale, built to last

05 / Working Out

Strong body,
strong mind.

Building systems is a marathon, not a sprint, so I train for both. This is my latest activity, pulled live from Strava.

// syncing with strava…

06 / Contact

Let's work
together.

Have an idea or a project? Drop me a line, I'd love to hear from you.

[email protected]