📍 Based in Frankfurt am Main, Germany
I turn operational and sensor data into explainable, governance-ready AI solutions
for aviation and logistics — from research prototypes to platform-oriented delivery.
I am a PhD Candidate in AI and Logistics and Research Assistant / Doctoral Researcher at Frankfurt University of Applied Sciences (Jan 2024 – Present). Within the Digital Testbed Air Cargo (DTAC) project, I develop applied AI and ML workflows for real aviation and logistics use cases — including SmartPouch-based cargo movement classification, ADS-B data processing, and explainable ML for airport ground operations.
My engineering direction is to grow into a strong AI Engineer / Data Platform Engineer, combining aviation domain experience with governed, production-oriented data and ML platforms.
I care about data and AI systems that are not only accurate, but also explainable, operable, and useful for domain experts under real operational constraints.
SoloLakehouse (SLH) is a self-hosted, cloud-neutral, governance-first lakehouse prototype. It connects aviation data standards, ML workflows, and metadata/lineage management — demonstrating how one engineer can design and operate an auditable data platform without depending on managed SaaS lakehouse vendors.
| Area | Design |
|---|---|
| Core stack | Apache Iceberg · Delta Lake · Trino · Dagster · MLflow · MinIO / SeaweedFS · Superset · OpenMetadata |
| Data path | Bronze → Silver → Gold → ML, with explicit governance and lineage boundaries |
| Governance posture | Architecture decisions documented for lineage, access control, and audit-friendly operations |
| Differentiation | Most open lakehouses optimize for scale. SLH optimizes for governed, explainable workflows where architecture decisions need to survive review |
RecordChat is an open-source, domain-specific AI assistant that makes IATA ONE Record easier to learn and explore. ONE Record spans specifications, ontology, REST API, JSON-LD payloads, and server implementations like NE:ONE — RecordChat shortens that learning path with grounded, source-cited answers instead of generic chatbot behavior.
| Area | Design |
|---|---|
| Core stack | FastAPI · Next.js · Qdrant · pluggable LLM / embedding providers · Docker Compose |
| Knowledge scope | ONE Record specs · ontology (classes, properties, relationships) · JSON-LD · NE:ONE server guidance |
| Domain tools | Ontology-aware retrieval & reranking · JSON-LD example generation · Mermaid diagrams for flows and relationships |
| Differentiation | Most AI chatbots optimize for fluency. RecordChat optimizes for traceability, helping developers and logistics teams understand ONE Record with auditable, source-linked answers |
Try asking: What is a LogisticsObject? · How do Shipment, Piece, and Waybill relate? · Generate a JSON-LD example for a Piece · How do I run NE:ONE locally?
Platform layers: sensor & operational data · open table formats · SQL & metadata · orchestration · ML lifecycle & explainability · GenAI/RAG APIs · observability · IaC
- HICL 2026 — Taxi-In Time Prediction for Airport Ground Operations Using Explainable Machine Learning
- LM25 2025, Milan — Comparing and Predicting Taxi-In Times for Passenger and Cargo Flights at Frankfurt Airport
- ATRS World Conference 2025, Hong Kong — An Enhanced System for Air Cargo Movement Detection Using Deep Learning and Explainable AI
Open to collaboration on aviation & logistics digitalization, research conversations, and AI / Data Platform Engineer roles at the intersection of applied ML, governed data workflows, and production-oriented platform engineering.