— Expertise in TensorFlow, Python, related libraries and OpenCV is a requirement.
— In-depth knowledge of Convolutional Neural Networks, including major network architectures (LSTM, inception, residual, GAN), knowledge of activation functions and their effects on network trainability and speed, experience in network training and tuning meta parameters.
— Experience in preparing and augmenting datasets for training, and synthesizing datasets using GAN.
— In-depth understanding of loss functions and ability to create custom loss functions.
— Understanding of GAN methodology and cross-entropy.
— Understanding of network sparsity, network pruning, and its effect on network speed and performance.
— Understanding of low bit depth networks, weight quantization and its effect on speed and trainability.
— Knowledge of performance metrics in object detection and classification, such as mAP and related.
— Familiarity with Nvidia GPU line of products and TensorRT software.
— Minimum experience: 3 years.
— Experience using JIRA or similar tracking / reporting systems is required.
— MS or BS in Math/Computer Science.
— Good written and spoken English (intermediate+).
Well be a plus
— CuDNN and ability to create custom network layers.
— Depth / 3D extraction using convolutional neural networks; anomaly detection using deep networks.
— Design, evaluate and improve computer vision solutions using Deep Learning and other techniques.
— Evaluate the performance of the solutions to ensure their quality.
— Maintain efficient regular communication with HQ; travel abroad is required.