Skip to content
View leonkoellerwirth-arch's full-sized avatar

Achievements

Achievement: Pull Sharkx2

Achievements

Achievement: Pull Sharkx2

Block or report leonkoellerwirth-arch

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
leonkoellerwirth-arch/README.md

Leon Köllerwirth Hlihel

Senior IT-Manager & Interim Leader · AI Governance & Agentic AI Operating Models · Enterprise Architecture in regulated environments (BaFin/DORA)

I've spent 20+ years building, leading, and orchestrating IT in regulated and critical environments — from greenfield corporate IT (Tönnies Group, 0→7 plants) through enterprise architecture, security and governance responsibility at a BaFin-regulated financial institution (Daimler Mobility AG), to platform governance with production AI agent pipelines (SMA Solar Technology).

Today I help organizations make AI agents governable: operating models, risk & control structures (EU AI Act, DORA), and the engineering patterns to match. The repositories below are working artifacts of that practice — governed agent pipelines, evaluator-tested prompts, and audit-first design.

Featured work

  • agentic-ai-governance-toolkit — lifecycle models, risk scoring, EU AI Act & DORA checklists, and a working evaluator for AI agents in regulated organizations
  • rag-approval-blueprint — the other half: not how to build a compliant RAG assistant, but how to get it approved — approval file, control catalogue with test procedures, deletion evidence, and a worked approval run (DORA/MaRisk)
  • local-agent-pipeline — a minimal, fully local multi-agent pipeline where auditability and human oversight are first-class features
  • razbiram-nlp — Bulgarian NLP content-enrichment engine (morphology, CEFR-style difficulty, glosses), with golden-set evaluated LLM prompts

Contact

🌐 leon-koellerwirth.com · 💼 LinkedIn

Pinned Loading

  1. razbiram-nlp razbiram-nlp Public

    Bulgarian NLP content-enrichment engine — morphology, CEFR-style difficulty, vocabulary, contextual glosses. Source-available work sample.

    Python 3

  2. local-agent-pipeline local-agent-pipeline Public

    A minimal, fully local multi-agent pipeline (Ollama) where auditability and human oversight are first-class: tamper-evident audit trail, action-space whitelist, human-in-the-loop gate, optional ext…

    Python 1

  3. razbiram-listen razbiram-listen Public

    Listen to Bulgarian audio and understand it word by word — local Whisper transcription in a synced karaoke reader. BYO-audio, private by design.

    Python 2

  4. agentic-ai-governance-toolkit agentic-ai-governance-toolkit Public

    Practical governance artifacts for AI agents in regulated organizations — lifecycle models, risk scoring, EU AI Act & DORA checklists, and a working evaluator.

    Python 1

  5. rag-approval-blueprint rag-approval-blueprint Public

    Everyone shows you how to build compliant RAG. This repo shows you how to get it approved. Freigabeakte, Kontrollkatalog mit Prüfhandlungen, Löschnachweis und ein durchgespielter Freigabelauf für B…

    TypeScript

  6. razbiram-anki razbiram-anki Public

    Bring your own Anki deck into the razbiram style — and into razbiram.com. Runs entirely in your browser: .apkg in, styled .apkg or CrowdAnki deck.json out. No install, no account, free.

    TypeScript