Freelancer profile translated to English.

Description

Enterprise AI Architect & AI Engineer for production-ready Artificial Intelligence, LLM, and automation solutions — from Agentic AI and RAG to robust enterprise architectures.

I help companies deploy AI where it creates measurable business value: in processes, customer workflows, operations, CRM workflows, and internal tools. No Proof-of-Concept theater — but solutions that run in production.

Focus & Tech Stack:

Production-ready AI solutions that integrate into existing enterprise systems — Azure, APIs, n8n, Python, TypeScript, LangGraph, Pydantic, HubSpot, and Salesforce-centric architectures. GenAI, LLM, RAG, Multi-Agent Systems, MLOps, and AI Governance.

Typical Projects:

Agentic AI / Multi-Agent Systems with LangGraph and Pydantic
LLM and RAG-based decision support for operations
Workflow Automation with n8n, APIs, and internal tools
AI Governance, architecture reviews, and enterprise readiness
Integration of AI into existing CRM, cloud, and process landscapes

Background:

14 years of enterprise IT, including 6 years as a Solution Architect / Enterprise Architect in a Fortune 500 environment. I combine technical depth with business understanding and clear stakeholder communication — from architecture and governance to implementation in real-world operations.

No PowerPoint prototypes. Production-ready AI solutions for real enterprise processes.

Do you have a specific AI or automation project in mind? Let's talk.

Languages: German / English

Industry field of expertise

Languages

German
Native or bilingual
English
Native or bilingual

Workplace preferences

Remote only

Primarily works remotely

Graphic Packaging International
Enterprise Architect | Head of AI
DIGITAL AND IT
September 2024 - Today (1 year and 9 months)
Berlin, Germany
Transformation portfolios are managed in Excel and reprioritized in quarterly offsites — budget follows the loudest stakeholder, not the greatest value contribution.

Solution: Composite scoring engine (4 weighted dimensions) + Capability maturity model (Process / People / Technology / Data) + AI recommendation layer with explainable reasoning + live portfolio dashboards. Read/write integration with the EAM catalog (LeanIX, ServiceNow APM, Ardoq). Deployment on-prem or private cloud.

Result:
• 12 live initiatives scored across 4 dimensions, portfolio €2.4M
• Scoping: 4–6 weeks workshop → minutes (AI-suggested)
• Reprioritization: Quarterly offsite → on every metadata change
• Allocation transparent: 67% committed / 21% planned / 12% free

Stack: Composite Scoring, Capability Radar, LLM recommendation layer, LeanIX / ServiceNow APM / Ardoq APIs.
AI Automation Cloud architect IT Architecture artificial intelligence LLM
Graphic Packaging International
Solution Architect | AI Solutions
DIGITAL AND IT
February 2022 - September 2024 (2 years and 7 months)
Berlin, Germany
SAP production data flows into Dev/QA/Training. Deterministic masking tools (SAP TDMS, Delphix, Informatica) cover ~95% of structured columns — gaps remain for free-text NOTES, unclassified Z-tables, and OCR scans. A general LLM over 100% of the data is not economically viable.

Solution: Fine-tuned German PII Redactor (LoRA adapter on Qwen2.5-1.5B) as a drop-in layer AFTER deterministic masking. Routes free text and unclassified columns to the model, delivering structured JSON (redacted text, entity classes, risk level, review flags) for 12 German entity types. Runs entirely on customer hardware (150 MB adapter), GDPR-compliant by design.

Result:
• Entity F1: 0.375 → 0.917 (2.4× better than baseline)
• JSON parse validity: 81% → 100%
• Risk level exact match: 0.50 → 0.94
• Training: ~90 min on a RunPod A40; inference on a consumer GPU

Stack: Qwen2.5-1.5B-Instruct, LoRA / PEFT / TRL, Pydantic-validated JSON, HuggingFace Hub, Apache 2.0.
artificial intelligence LLM Agentic AI AI Automation
Graphic Packaging International
Information Technology System Administrator
DIGITAL AND IT
January 2012 - August 2020 (8 years and 7 months)
Berlin, Germany
Led the migration of 2,500 mailboxes across two Microsoft 365 tenants, covering 30 locations and 5 languages. Took the technical lead, built the project plan, delegated to external contractors, and coordinated rollout across time zones until every mailbox was moved.
Kubernetes Microsoft Azure Python Integration Engineer Software Development