Applied AI / Data Science Engineer | Computer Vision | Industrial AI

I work at the intersection of data science, machine learning, computer vision, and deployment-oriented AI systems. My strongest experience comes from industrial AI projects, where I developed and evaluated anomaly detection pipelines for real-world inspection data, including a weld-seam surface inspection workflow that achieved AUROC 0.966 and was benchmarked against commercial HALCON software.

I am currently building my portfolio toward data science, operational analytics, LLM/RAG systems, and production-aware computer vision applications.

• Turning raw data into structured analysis, model inputs, and decision-ready insights
• Building reproducible workflows for data preparation, model evaluation, and benchmarking
• Developing computer vision pipelines for defect detection, anomaly localization, and quality inspection
• Designing AI systems with deployment in mind: Docker, APIs, cloud storage, logging, and monitoring
• Expanding into LLM/RAG systems for knowledge retrieval, automation, and decision support

SQL-driven analysis of marketplace order data to understand delivery performance, customer satisfaction, and operational bottlenecks.

What this project shows

• End-to-end analysis using SQL and Python
• Root-cause investigation of fulfilment delays and late deliveries
• Hypothesis testing and A/B-style analysis to estimate business impact
• Clear communication of insights through metrics, visualizations, and recommendations

SQL DuckDB Python pandas SciPy A/B testing Hypothesis testing Operational analytics

Methodology and case-study repository based on my Master's thesis and industrial work at Automation W+R GmbH.
The original code and data are proprietary, so this repository documents the architecture, experimental design, evaluation strategy, and deployment-oriented thinking behind the project.

What this project shows

• Computer vision pipeline for industrial 2D/3D weld-seam surface data
• Autoencoder-based unsupervised anomaly detection and defect localization
• Benchmarking against HALCON with AUROC 0.966
• Evaluation using AUROC, precision, recall, F1-score, SSIM, MSE, and failure-case analysis
• Deployment-oriented workflow using Docker, AWS EC2/S3, REST API concepts, Git/GitHub Actions, and Python-based logging

Computer Vision Anomaly Detection OpenCV Autoencoders TensorFlow/PyTorch HALCON Docker AWS

NDA-safe methodology repository based on my Master's thesis at Automation W+R GmbH.
Documents the problem setup, dataset preparation, autoencoder-based anomaly detection, HALCON benchmarking, failure analysis, and deployment-oriented design for industrial 2D/3D weld-seam inspection.

View repository

I am currently building a retrieval-augmented generation project for domain-specific question answering. The goal is to create a reliable assistant that retrieves relevant context, answers with grounded information, and evaluates response quality using a curated question set.

Current focus

• Document ingestion and chunking
• Embeddings and vector search
• Retrieval-augmented generation pipeline
• Response quality evaluation
• Docker-based local deployment

LLM RAG Embeddings Vector Search LangChain Python Docker

• 🎓 M.Eng. Engineering Sciences – Mechatronics, Technische Hochschule Rosenheim
• 🏢 Machine Learning / Computer Vision Engineer, Automation W+R GmbH, Munich
• ⚙️ Experience with industrial inspection systems, 2D/3D surface data, anomaly detection, and engineering validation workflows
• 📍 Based in Rosenheim, Germany — open to onsite/remote/hybrid opportunities in EU
• 🌐 English C1 · German B1, actively improving

I am building a portfolio that connects my industrial AI background with broader data science and modern AI engineering:

• Data science and operational analytics for decision support
• Computer vision systems that can move from prototype to deployment
• LLM/RAG applications with evaluation and reliability in mind
• Production-aware AI workflows with reproducibility, logging, and deployment planning