16.10.2026 aktualisiert
100 % verfügbar
Freelance Senior AI Engineer & Software Engineer | Computer Vision & Medical Imaging Specialist
Pulheim, Deutschland
Weltweit
Master of Engineering - Software Engineering, Northwestern Polytechnical University, Xi'an, ChinaÜber mich
Freelance AI Engineer & Software Engineer mit Fokus auf Computer Vision & Medical Imaging. Expert in Python/PyTorch (MEA-UNet, Dice 0.8185) & C++/OpenCV Integration (Nissan-Produktionslinien). Ansässig in Pulheim. Fließend in Deutsch (C1) und Englisch (C1).
Skills
Künstliche IntelligenzComputer VisionC++PythonMATLABMachine LearningOpenCVImagingPyTorchMachine Learning OperationsFeature Extraction
AI and Computer Vision
Expertise in image segmentation, classification, ML prototypes, and model benchmarking using Python, PyTorch, TensorFlow, and OpenCV. Development of production-grade, reproducible ML pipelines including MEA-UNet for retinal fundus vessel segmentation with multi-scale and edge-aware feature extraction.
Data and Sensor Pipelines
Proficiency in preprocessing, feature engineering, radar and THz signal analysis, and reproducible evaluation using Python and MATLAB. Experience building THz signal-processing workflows and Python ML pipelines for gesture recognition from millimetre-wave radar measurements.
C++ and System Integration
Strong background in C/C++, MFC, TCP/IP middleware, SQL Server interfaces, machine-vision tools, and backend development. Delivered production-ready machine-vision systems deployed on Nissan production lines using C++/MFC TCP/IP middleware and OpenCV ROI and template-selection tools.
Project Delivery and QA
Full-lifecycle expertise covering requirements engineering, implementation, integration testing, debugging, and releases using Git/GitHub. Led a four-person team from requirements through delivery on industrial machine-vision projects.
Medical Image Segmentation
Specialization in Medical Image Segmentation, including Fundus and MRI/CT pipelines, with demonstrated results outperforming U-Net, Attention U-Net, and CBAM U-Net on Dice and IoU metrics across multiple datasets.
THz Sensor Machine Learning
Experience in THz signal classification using neural networks, achieving over 80% accuracy in classifying nine materials and detecting signals up to 600 GHz at 1.2 m distance.
Backend and Database Development
Development of C backend modules and SQL Server interfaces for banking workflows, supporting QA, releases, and defect resolution.
IT Support and Hospital Systems
Support of HIS, EMR, PACS, and LIS platforms, resolving software, endpoint, and network issues for hospital users.
Erfahrung
Berufserfahrung5 - 10 Jahre
Freelancer seit2022
Sprachen
ChinesischMutterspracheDeutschverhandlungssicherEnglischverhandlungssicher
Projekthistorie
Developing production-grade, reproducible ML pipelines. Built a reproducible PyTorch pipeline for image preprocessing, training, testing, visualization, and evaluation. Developed MEA-UNet for retinal vessel segmentation with multi-scale and edge-aware feature extraction. Outperformed U-Net, Attention U-Net, and CBAM U-Net on Dice and IoU across three datasets; best Dice: 0.8185 on HRF. Presented the work orally at the Optica ImageSense Congress 2026.
Built a Python ML pipeline for gesture recognition from millimeter-wave radar measurements. Performed radar-signal feature extraction and trained classification models for contactless sensing. Improved model robustness through feature engineering and systematic evaluation.
Built THz signal-processing workflows, including removal of water-vapor absorption artefacts. Classified nine materials with over 80% accuracy using a neural network; detected signals up to 600 GHz at 1.2 m.