Dr. Paul Dragan

I co-founded Kladero GmbH to help organizations design, build, deploy, and operate practical AI and software solutions. I work hands-on across AI and machine learning, software architecture, backend and distributed systems, cloud platforms, Kubernetes, and DevOps, while leading company strategy, product direction, customer relationships, and business development. I also lead the development of vera[tag], Kladero’s SaaS platform for managing and displaying AI-content disclosures for EU AI Act compliance.

Researched AI, machine learning, distributed and self-adaptive systems, data protection, and regulation. Developed reinforcement-learning and constraint-optimization approaches for coordinating adaptive systems and preventing consent violations at runtime. Led a work package in the EU Horizon project FogProtect, coordinated technical activities, taught software engineering, and supervised student projects and theses.

Optimized and deployed large-scale CNNs used in automotive radar applications on resource-constrained and heterogeneous embedded hardware. Evaluated model architectures and profiled compute, memory, and I/O bottlenecks. Developed compiler-driven tooling for model transformation, deployment, and execution across CPUs and specialized accelerators. Technologies included C++, Python, TensorFlow, ONNX, TVM, CUDA, ARM, and TI SoCs.

Developed compiler technology to run machine-learning applications efficiently on ARM processors. Developed computer-vision algorithms for semantic segmentation and road-marking detection to improve traffic-scene understanding. Worked with C++, OpenCV, PyTorch, NVIDIA and ARM platforms, and embedded systems.

Conducted BSc thesis research at the Delft Center for Systems and Control on robust object matching across multiple camera views. Designed a CNN-based multi-view object-matching algorithm for robotic perception under changing viewpoints using Python, TensorFlow, ROS, OpenCV, and Blender.

Developed machine-learning and computer-vision algorithms for road and free-space detection to support environmental perception in automotive embedded systems. Worked with Python, C++, TensorFlow, and OpenCV.

Researched and developed assistive-robotics applications for monitoring and safety of at-risk people using UAVs, mobile robots, and robotic manipulators within the Robotics and Nonlinear Control research group. Worked with C++, Python, TensorFlow, and ROS.

Implemented a factory-automation system to replace manual consumables tracking, estimate wear, and support timely replacement. Developed C and C++ software with NI LabWindows/CVI and integrated measurement hardware and sensors for industrial process monitoring.

PhD in Software Engineering, awarded magna cum laude. Research focused on AI, machine learning, distributed and self-adaptive systems, data protection, and regulation. Dissertation: “Preventing consent violations due to self-adaptation at runtime.” Supervised by Prof. Dr. Klaus Pohl.

MSc in Control Engineering, final grade 99.60%. Focused on advanced process control, AI, mathematical optimization, distributed systems, and computer vision. Thesis: “Traffic scene semantic segmentation from depth images with convolutional neural networks.” Supervised by Assoc. Prof. Dr.-Ing. Mihai Negru.

BSc in Control Engineering, final grade 97.10%. Focused on control engineering, distributed and real-time systems, computer vision, AI, and robotics. Thesis: “Object matching with convolutional neural networks.” Supervised by Prof. Dr.-Ing. Lucian Busoniu and Prof. Dr.-Ing. Robert Babuska.