Junaid

Computer Vision engineer

Kicks • Fulltime

Oct 2025 - Present • 7 mos

Model Development & Training Custom Architecture Design: "Architected and fine-tuned state-of-the-art Deep Learning models (YOLO, Faster R-CNN, EfficientNet) for specialized object detection tasks." Image Segmentation: "Implemented advanced semantic and instance segmentation techniques using Mask R-CNN and U-Net for medical imaging and satellite data." Transfer Learning: "Leveraged transfer learning on pre-trained models (ResNet, VGG) to achieve 95%+ accuracy on niche datasets with limited samples." Deployment & Optimization Edge AI Deployment: "Optimized deep learning models for edge devices (Jetson Nano, Raspberry Pi) using TensorRT and OpenVINO, reducing latency by 40%." Cloud Integration: "Deployed scalable inference APIs using Docker and FastAPI, hosted on AWS (SageMaker) and Google Cloud Vertex AI." Model Compression: "Applied quantization and pruning techniques to compress large models for mobile integration without significant loss in precision."