At Haensel AMS GmbH, I migrated daily data pipelines for 10 clients from event-driven AWS services to Apache Airflow on Kubernetes, cutting time to data availability from 8 hours to 2. I’ve since shaped the multi-client platform’s data quality and SLA guardrails, automated routine failure recovery, and introduced tests to a codebase that had none.
I build lakehouse layers on S3 with Apache Iceberg and Delta Lake, and develop reusable dbt data products on Redshift and Snowflake. I also improved performance on the heaviest analytical workloads, reducing query times by 55% with a matching drop in storage spend.
For a production LLM assistant for accounting and tax analysis, I owned the data side: document chunking, embeddings, OpenSearch indexing, retrieval, and output evaluation. I use AI coding agents in my everyday engineering work and share where human review is needed.
Earlier, at Fraunhofer IISB, I ran simulations of photoresist behaviour in EUV lithography and used graph neural networks to label a 63k-pattern dataset for my master thesis. At Accenture Solutions Pvt. Ltd., I built Java and Spring Boot loyalty-programme services and supported production systems.

