At the intersection of embedded logic, real-time operating systems, AI orchestration, and grid-scale infrastructure. My work translates high-stakes technical complexity into systems that are predictable, legible, and deployable in safety-critical environments.
I do not treat them as isolated domains. From RTOS scheduling to AI orchestration, each layer has to be checkable before the one above it is allowed to depend on it.
ITk Fachspezialist — Digitisation of high-voltage assets
DB InfraGO AG · Aug 2024 – present · Frankfurt
Digitalisation of railway traction HV grids, IT/OT convergence, and KRITIS-aligned cybersecurity governance for mission-critical rail infrastructure.
PreviouslyIndustrial Engineering Intern — High-voltage maintenance
DB Fahrzeuginstandhaltung GmbH & DB Netz AG · Jun 2022 – Sep 2024
Lifecycle management of traction power substations, asset condition monitoring, and predictive maintenance.
Grouped by domain rather than by tool. Select a domain to filter the work registry underneath it.
Modelling, coordinating and dispatching electrical assets — from traction substations to distributed energy resources.
Getting field devices to speak to software without losing determinism, and keeping the boundary between the two defensible.
Embedding governing equations and threat models into learned components so their outputs stay physically admissible.
Turning raw sensor returns into geometry a controller can act on, with the calibration maths done properly.
Letting language models reach real actuators and real ledgers without giving up an audit trail or a kill switch.
The delivery substrate: typed services, native extensions, real-time browsers surfaces, and pipelines that publish.
Interactive simulator for the cross-domain CIM + ThreMA ontology, PINN solvers, RL security agents and IEEE 9-bus cyber-physical validation.
DER middleware and control surface: protocol ingest, MILP dispatch, anomaly detection, and an MPC forecast loop.
Distributed energy resource integration middleware focused on verifiable coordination and grid-aware execution.
Governance layer for AI agents acting on financial systems: deterministic policy engine, signed action tokens, hash-chained audit log.
LiDAR-to-camera projection with SE(3) extrinsics, pinhole intrinsics, z-buffer occlusion handling and a KITTI calibration loader.
AI-native middleware for human-to-model orchestration in safety-critical, physics-informed environments.
Physics-informed intelligence does not stop at pattern recognition. It constrains learning with the same governing equations that define the physical system.
Real-time surrogate models for optimal power flow and inverter control — systems that are not merely intelligent, but operationally trustworthy under physical constraints.
The first systematic integration of the Common Information Model (CIM) with the ThreMA cybersecurity framework: unified semantic representations connecting physical power components with vulnerabilities and protective measures. Validated on an enhanced IEEE 9-bus system.
Consultation deposits, advisory retainers, and invoice payments through encrypted Stripe checkout. Cards, mobile wallets, and SEPA accepted — with an instant receipt.
A focused, high-density working session on grid intelligence, cyber-physical systems, or physics-informed AI.
A dedicated advisory and engineering block with priority access and guaranteed turnaround. Cancel anytime.
Enter exactly what you owe — from €5 to €5,000 — and pay with the method you prefer.
AI-native middleware, smart-grid operating systems, embedded control platforms, robotics, or research collaboration. The repositories are private; access is granted on request.