25.11.2025 aktualisiert
Premiumkunde
100 % verfügbarComputer Vision & Deep Learning Engineer
Maintal, Deutschland Master of Science, International Automotive Engineering
Über mich
Specialized in bridging gap research & production. Optimize deep learning models for real-time Edge/Cloud. With 7+ years exp, I engineer high-performance C++/Python pipelines using TensorRT & custom CUDA kernels. I turn experimental DL models into fast, reliable, industrial-grade production systems.
Skills
Künstliche IntelligenzAmazon Web ServicesComputer VisionC++CmakeCUDAComputerprogrammierungLinuxProgrammierwerkzeugeFfmpegPythonNumpyObjekterkennungOpencvTensorflowSQLUrduVideoverarbeitungNetwork RoutersDatenverarbeitungPytorchHigh Efficiency Video Coding (HEVC)Deep LearningKerasGitSensorikPandasMatplotlibScikit-learnHardware AccelerationMachine Learning OperationsDSGVOLidarDockerProgramming LanguagesData Generation
Senior Computer Vision & Deep Learning Engineer (C++ / CUDA / Python)
Specialized in building high-performance, production-grade computer vision pipelines. I bridge the gap between Deep Learning research and real-time industrial deployment using Hardware Acceleration and Model Optimization.
Core Competencies:
- Deep Learning & AI: Object Detection (YOLO), Instance Segmentation (Mask RCNN), Tracking (SORT), OCR, 6D Pose Estimation.
- High-Performance Computing: Custom CUDA Kernels, TensorRT Optimization, FFmpeg (libav) video processing, GPU Acceleration.
- Programming: Expert in Python and C++ (Modern C++, STL).
- MLOps & Data: Synthetic Data Generation (Blender/Unity pipelines), Managing TB-scale datasets, Docker, Git.
- Hardware: Integration with Industrial Routers (Teltonika), Sensors (LiDAR, RGB-D), and Edge Devices.
Tech Stack:
- Frameworks: PyTorch, TensorFlow, Keras, TensorRT, OpenCV.
- Languages: Python, C++, CUDA C++, SQL.
- Tools: Linux, Docker, AWS, Git, CMake.
- Languages: English (Native), German (Proficient/Verhandlungssicher), Urdu (Native).
Sprachen
DeutschverhandlungssicherEnglischMuttersprache
Projekthistorie
High-Performance Video Analysis Engine
- Engineered a hardware-accelerated media player in C++ using libav (FFmpeg) and SDL3 for high-speed H.265 video processing.
- Authored custom CUDA kernels for motion detection and background subtraction, achieving 1000+ FPS on GPU.
- Implemented ring buffers and shared hardware CUDA contexts for low-latency packet navigation.
- Developed tools for managing and verifying terabyte-scale video datasets.
GDPR-Compliant Traffic Monitoring AI
- Trained and deployed YOLO-based object detection models optimized with TensorRT for high-throughput license plate recognition.
- Deployed Mask R-CNN instance segmentation for GDPR-compliant blurring of people and vehicles.
- Implemented Deep Learning-based tracking (SORT) and OCR for vehicle identification.
- Integrated AI models with industrial hardware (Teltonika routers) via webhooks.
Synthetic Data Pipeline for 6D Pose Estimation
- Generated a large-scale synthetic dataset (6 million images) to train object detection models.
- Designed End-to-End Deep CNNs for 6D Object Pose Estimation using RGB-D data.
- Applied instance segmentation for defect detection in wind turbines and agricultural applications.
- Performed sensor fusion of RGB, Thermal, and LiDAR data for workspace monitoring.