since 01/2024 · Robert Bosch GmbH
Cloud, Data & AI Solutions Architect
Schwieberdingen, Germany
Designed and introduced an end-to-end Data & AI platform architecture on the Databricks Lakehouse across hybrid and public cloud, serving as trusted technical advisor to cross-functional teams and delivering production-grade data engineering and analytics at scale in an industrial and manufacturing context. Solution architect across 4+ integrated production solutions:
- Architected a rich-code multi-agent orchestration platform (LangGraph, Google ADK) improving design quality and cutting layout cycle time by 26% through deploying autonomous, dynamic agentic systems that adapt their topology to a particular CAE task — optimising layout, geometry and design of automotive ECU components — freeing engineers from repetitive tasks.
- Delivered an end-to-end AI- and data-driven system — from data pipelines through a semantic knowledge layer (explainable/interpretable AI and knowledge graphs) to a self-service interface — democratising knowledge sharing across engineering processes for 10+ product and domain teams and accelerating process improvement.
- Architected and prototyped a hybrid-cloud infrastructure framework running virtual load simulations of ECU designs on on-site HPC systems, with an AI-optimised runtime that seamlessly bursts demand spikes to cloud HPC nodes — increasing simulation throughput by a factor of 5 for smaller jobs while controlling compute cost.
- Designed an event-based, near-real-time streaming data pipeline (Kafka, Flink, Delta Lake) giving test engineers near real-time access to long-running environmental test data, powering early-warning and anomaly-detection systems that reduced time-to-detection by up to 2+ weeks, potentially saving over €1M per year and lab.
03/2022 — 12/2023 · Robert Bosch GmbH
Senior Data Architect
Schwieberdingen, Germany
Introduced modern data modelling and data-mesh methodology to enable data-driven decision-making, delivering high-fidelity curated data products and unified digital component CVs across the complete ECU product-development lifecycle. Led and advised a global, interdisciplinary team of 12 data-modelling domain experts and data engineers to build a unified data model — from physical sources up to the target business layer.
- Designed a data product built on graph-based semantic technologies and a federated data layer spanning the full chain from requirements through design to manufacturing capabilities, enabling design-to-shop-floor matching for cost optimisation across the ECU product portfolio. Lowered barriers between engineering, manufacturing & logistics — reducing lead & development time for new design improvements by a factor of 2.
- Built a showcase demonstrating Databricks Delta combined with LakeFS for full data traceability and baselining, driving evaluation and adoption of the platform across the division, and supported the introduction of a central tool applying a unified ontology for rapid feasibility checks of new designs against existing product digital CVs — cutting feasibility-check turnaround by 40%.
01/2018 — 01/2022 · Robert Bosch GmbH
Data & ML Engineer
Reutlingen, Germany
Built and productionised machine-learning systems for manufacturing — from data pipelines through model training, deployment and monitoring — owning the full MLOps lifecycle on edge and cloud to improve quality and reduce cost on the production line.
- Developed RCAS, a Spark-based root-cause recommender that predicts failures in the pin-and-connector press-in process by detecting distinctive patterns in measurement data and cross-checking against known 8D reports — surfacing root causes over 50% faster with 2x higher accuracy.
- Built a deep-learning system supporting automated optical inspection (AOI) of solder joints, scoring images flagged as pseudo-failures and categorising the potential failure type — reducing manually reviewed images by a factor of 10 and outperforming human operators while cutting operator fatigue. Shipped to an edge device on the production line (MES-controlled) with shadow-mode deployment, model-drift detection and automated retraining triggers for new packages/failure types.
09/2015 — 12/2017 · Robert Bosch GmbH
Connected Industries Data Scientist
Reutlingen, Germany
Built and evolved an early data and AI platform architecture on hybrid and public cloud, laying the foundation for data-driven analytics across industrial and manufacturing operations.
- Designed an SMT process monitor that integrates and harmonises data across the full surface-mount value stream — solder-paste application, pick-and-place, reflow soldering, automated optical inspection (AOI) and in-circuit test (ICT) — consolidating cycle-time data from every machine into a single unified model on Apache Spark. By correlating data along the whole value stream, the system flags defects that would only surface at AOI already at the solder-paste stage, letting engineers scrap the bare panel before expensive components are placed and soldered onto it — reducing material scrap cost by 12%.
- Built a demand-forecasting system predicting customer order volumes across multiple horizons (1-, 3- and 12-month windows) and modelling each customer's systematic forecast bias — learning empirically where customers over- or under-estimate future demand — to plan production volumes more accurately and optimise supplier material orders, keeping inventories low while improving demand-forecast accuracy by 60% and reducing inventory costs by 15%.
09/2013 — 08/2015 · Universiteit Leiden
Postdoc/Research Associate
Netherlands
Modelling and computation of non-linear partial differential equations on HPC supercomputing clusters.