Alongside my doctoral research, I take on a select number of consulting mandates. My work bridges the gap between traditional approaches and AI-augmented systems in modelling and forecasting, with a focus on environmental risk and on quantifying the uncertainty of extreme events. I advise public authorities, industry and decision-makers, from the initial analysis to the system in operation.
Federal Waterways Engineering and Research Institute (BAW)
Aim Forecasts orders of magnitude faster than the numerical model, just as accurate, in daily use at BAW.
In KIOM, BAW is working on machine-learning surrogates for its hydrodynamic models, meant for its day-to-day forecasts. As lead developer and consultant, I help build them, bringing in a graph-neural-network approach I tested and published in research in 2026.
Hamburg Port Authority
Result A machine-learning fallback, running live in operation.
The Port of Hamburg's operational current forecast needed a fallback in case the external forecast it builds on fails. As project lead, I introduced a data-driven water-level forecast and integrated it into the operational model, where it takes over when needed.
Hamburg Port Authority
Result Uptime rose to about 99.5%, with roughly 30% less code to maintain.
OpCIS, the Port of Hamburg's real-time current forecasting system, has to run reliably around the clock. I kept it running and improved it, maintaining its databases and related systems, and streamlining the codebase.
Hamburg Port Authority
Result A new atlas with fewer spatial errors, now used by pilots and simulators.
Pilots and ship-handling simulators rely on a current atlas of the Port of Hamburg that was due for an update. I rebuilt it for a range of conditions, including extreme events, simulating five scenarios across the port and deriving continuous current maps from them.
Hamburg Port Authority
Result The client could inspect the effects on manoeuvring and sedimentation directly.
The port's planned western extension could change how ships manoeuvre and cause sedimentation. To make both effects visible for the construction plan, I set up, ran and analysed hydrodynamic simulations of several scenarios and turned them into animations.
Confidential client
Result The study pinpointed bottlenecks in vessel traffic and port operations.
A port needed a capacity assessment. To find out how vessels really use the port, I built a Python package that derives transit routes and operational patterns from AIS data.
Confidential client
Result Volume reports in about 30 minutes instead of a day.
An infrastructure project needed volume changes from frequent seabed surveys, with requirements that changed at short notice. We jointly analysed the multi-beam surveys with analysis software I developed as part of the project, adapting it as the requirements moved.