Principal AI Engineer

12/2023 – offen

Engineered a production multi-agent system (Amazon Bedrock + LangGraph) with 5 specialised agents, achieving 94% extraction accuracy — validated through continuous evaluation pipelines including Ragas benchmarking, hallucination detection, and prompt versioning — enabling end-to-end KYC/AML automation.
▪ Built a hybrid RAG architecture (Pinecone + OpenSearch) indexing 500+ pages of compliance policies, reducing false positives by 45%.
▪ Implemented an enterprise integration layer using Azure Databricks (Delta Lake) and optimised database connectors to orchestrate bidirectional data flows between real-time vector processing systems and core banking ledgers.
▪ Developed Spring Boot microservices (Java 17) as an AI API Gateway handling 2M+ transactions/day; reduced end-to-end latency by 40% using rate limiting and circuit breakers.
▪ Built event-driven integration pipelines using Spring Kafka processing 500K+ events/day with exactly-once semantics, achieving fault-tolerant real-time synchronisation between autonomous AI agents and enterprise financial state machines.
▪ Contributed to an 18-month AI transformation roadmap for enterprise GenAI adoption, supporting the establishment of an AI governance framework aligned with EU AI Act and BaFin requirements — part of a platform initiative that secured €6M investment.
▪ Implemented defence-in-depth security (OAuth2, prompt injection prevention, PII masking), ensuring GDPR and PCI DSS compliance.
▪ Deployed containerised services on Kubernetes with auto-scaling and CI/CD pipelines, reducing deployment cycles from 2 weeks to 4 hours and cutting infrastructure costs by 35%.

Bankwesen (allg.)