Beschreibung

I am a Lead AI Engineer with over ten years of data engineering experience and three years shipping production LLM applications. I help teams turn fragile prototypes into reliable AI architectures by owning the technical roadmap and establishing rigorous evaluation practices. My focus is on driving measurable business impact through scalable systems. Recently, I deployed a multi agent workflow that cut financial reconciliation time from ten days to just two hours, drawing on my background of scaling platforms to process 100 million daily events at 99.96 percent uptime.

Sprachen

Polnisch
Muttersprachlich oder zweisprachig
Englisch
Verhandlungssicher
Deutsch
Grundkenntnisse

Arbeitsortpräferenzen

Vor Ort möglich

Berlin (bis zu 50 km)

Self-employed
Freelance AI Engineer
BERATUNG & AUDITS
November 2025 - Heute (7 Monate)
Berlin, Germany
Reduced GDPR compliance review time from 7 days to 1 day for a data platform client by building an autonomous LLM classification system to catalog PII across 5 TB of unstructured enterprise data
Discovered and validated 12 undocumented vulnerabilities in Python dependencies by building a multi stage LLM agent pipeline with taint analysis and sandboxed Docker execution
Reduced manual screening time by 40% for an HR tech scaleup by building a dual-vector-store semantic matching system using pgvector and Weaviate
Built a headless browser abstraction layer producing structured ARIA JSON from bot-protected websites via Camoufox for subprocess-based LLM web data extraction
Built a custom LLM evaluation framework using inspect-ai with deterministic scoring and sandboxed code execution, benchmarking performance across 16 capability domains
LLM AI Agent Fullstack MLOps AI Automation
Babbel GmbH
Head of Data Engineering
BILDUNG & E-LEARNING
Januar 2022 - Oktober 2025 (3 Jahre und 9 Monate)
Berlin, Germany
•Decreased financial report reconciliation time from 10 days to 2 hours by deploying a multi-agent LLM workflow with deterministic validation guardrails and a React human-in-the-loop review UI
Reduced marketing copy production time from 4 hours to under 10 minutes per asset by building a Generative AI content engine with custom prompt pipelines
Reduced data processing costs by €18,000/month by migrating 81 data pipelines to Databricks and dbt
Reduced proof-of-concept delivery from 14 days to 2 days by adopting AI-assisted development (Cursor, Copilot, Claude Code) and upskilling teams on LLM integration paterns
Refactored a 21,000-line Python codebase into modular packages with CI/CD quality gates, reducing PR review time from 48 hours to under 4 hours
Eliminated 60 hours of monthly engineering work by automating GDPR user-deletion across 80+ datasets with a centralized service
Architected a real-time payment analytics layer handling 10M+ daily transactions from Stripe, Adyen, and PayPal at 99.9% uptime
Built an Internal Developer Platform with Infrastructure-as-Code, providing 11 product teams self-service storage and compute provisioning
Prevented 45 downstream reporting outages per quarter by implementing JSON schema validation across 7 cross-functional teams
AI Automation AI Agent LLM Data Governance Data Engineer
Babbel GmbH
Engineering Manager
BILDUNG & E-LEARNING
Januar 2019 - Januar 2022 (3 Jahre)
Berlin, Deutschland
Increased targeted campaign effectiveness by 15% (measured by CTR) by developing an ML-powered customer segmentation model
Raised deployment frequency from monthly to weekly and sprint delivery accuracy from 48% to 92% by introducing Scrum and CI/CD
Built the Data Engineering team from 0 to 4 members, achieving full sprint capacity within 6 months
Reduced critical outages from 5/month to 0 by implementing monitoring and incident response processes
Reduced MTTR from 4 hours to 42 minutes by standardizing incident runbooks and alerting thresholds
Designed a distributed payment processing engine handling 500+ transactions per second with idempotency guarantees
Built a data ingestion API sustaining 50,000 events/second with zero data loss during traffic spikes by implementing Kinesis with backpressure handling
Data Engineer MLOps Python Snowflake Databricks