As a Computer Vision Engineer with extensive expertise in OpenCV, PyTorch, TensorFlow, MediaPipe, and Nvidia SDK, I excel in bringing cutting-edge vision technologies from concept to reality. My work encompasses advanced pose estimation methods (VideoPose3d, Movenet, BlazePose, OpenPose), object detection (YOLO family, DETR, DINO, EfficientDet , MobileNet), and tracking (Lucas Kanade, SiamMask, Deep Sort), coupled with skills in perspective transformation and 3D geometry. I am adept at fine-tuning state-of-the-art models and crafting algorithms from scratch, with a focus on model optimization techniques like quantization with TensorRT. My experience spans implementing solutions on diverse platforms, from AWS servers to edge devices like Jetson Nano and Raspberry Pi. A unique aspect of my work is creating synthetic data using Unity3D, Nvidia omniverse, Unreal, Blender particularly valuable in scenarios with limited or unbalanced datasets. Also expert in training synthetic data model and close real and synthetic domain gap! As well as create special diffusion models to enhance the same. If yo want to distill your pre-trained models in some other models. I can do it! If your model is tough to train for your use case just bring it on! I am also skilled in setting up and tinkering with Linux systems for research, ensuring seamless integration with NVIDIA GPUs and deep learning frameworks. Driven by a passion for innovation in Computer Vision, I am committed to developing intelligent algorithms that deliver tangible value in real-world applications.As a Computer Vision Engineer with extensive expertise in OpenCV, PyTorch, TensorFlow, MediaPipe, and Nvidia SDK, I excel in bringing cutting-edge vision technologies from concept to reality. My work encompasses advanced pose estimation methods (VideoPose3d, Movenet, BlazePose, OpenPose), object detection (YOLO family, DETR, DINO, EfficientDet , MobileNet), and tracking (Lucas Kanade, SiamMask, Deep Sort), coupled with skills in perspective transformation and 3D geometry. I am adept at fine-tuning state-of-the-art models and crafting algorithms from scratch, with a focus on model optimization techniques like quantization with TensorRT. My experience spans implementing solutions on diverse platforms, from AWS servers to edge devices like Jetson Nano and Raspberry Pi. A unique aspect of my work is creating synthetic data using Unity3D, Nvidia omniverse, Unreal, Blender particularly valuable in scenarios with limited or unbalanced datasets. Also expert in training synthetic data model and close real and synthetic domain gap! As well as create special diffusion models to enhance the same. If you want to distill your pre-trained models into some other models. I can do it! If your model is tough to train for your use case just bring it on! I am also skilled in setting up and tinkering with Linux systems for research, ensuring seamless integration with NVIDIA GPUs and deep learning frameworks. Driven by a passion for innovation in Computer Vision, I am committed to developing intelligent algorithms that deliver tangible value in real-world applications.