Syed Zoraiz A.

Need a Computer Vision Engineer who can turn image or video data into a reliable production system? I build real-time computer vision and machine learning pipelines for object detection, tracking, segmentation, pose estimation, OCR/ANPR, industrial inspection, and video analytics. My core stack is Python, PyTorch, YOLO, OpenCV, ONNX Runtime, TensorRT, CUDA, FastAPI, and Docker. 3+ years in Computer Vision and ML 100% Upwork Job Success Score with consistent 5-star feedback $20K+ delivered across Upwork and direct engagements Selected results: * Built and evaluated an automotive component-inspection system using 20,000+ images * Developed real-time posture and movement analysis with custom keypoint logic, calibration, temporal smoothing, and CPU-friendly inference * Built multi-camera vehicle detection, persistent tracking, ANPR/OCR, and event logging * Optimized a background-removal pipeline from 2+ seconds to 0.8 seconds and reduced 4K image-saving time from about 13 seconds to 2 seconds I can help with: Computer Vision and Deep Learning * Object detection, classification, instance/semantic segmentation, pose estimation, and tracking * YOLO training, transfer learning, dataset preparation, augmentation, evaluation, and error analysis * RTSP, CCTV, webcam, multi-camera, and recorded-video analytics * OCR, ANPR/ALPR, document vision, barcode, label, and product recognition * Industrial inspection, defect detection, image registration, measurement, and anomaly detection * Research-paper implementation and custom algorithm development Inference Optimization and Deployment * PyTorch to ONNX conversion and ONNX Runtime/TensorRT acceleration * FP16/INT8 quantization, batching, profiling, memory reduction, and CPU/GPU optimization * FastAPI/REST APIs, Docker, Redis workers, databases, dashboards, and Electron integration * Using NVIDIA Triton Inference Server or NVIDIA Deepstream for pipelines and delivery * Deployment on Windows, Linux, AWS, RunPod, local servers, and edge hardware You receive tested, maintainable code, clear documentation, and a system designed around your target accuracy, FPS, latency, privacy, hardware, and operating environment. Send me your sample images/video, current code or model, target hardware, and required accuracy/FPS. I’ll recommend the most practical route from prototype to deployment.