We are looking for a Staff/Senior Machine Learning Engineer to take our Edge AI team to the next level. This team works within the larger ASIC Design team that develops and delivers toolchains that allow customers and partners to compile AI models to run optimally on resource-constraint environments like Neural Processing Units (NPU).
Responsibilities:
Research, experiment and develop new Machine Learning packages for efficiently running CNN models on NPU for various EdgeAI applications
Enhance, maintain and support existing toolchain features, model training pipelines, model zoo and modify model architectures for NPU optimality.
Develop python packages for supporting toolchain in various phases of compiler workflow.
Serve as SME within the team to support ML queries for internal & external users.
Keep management up to speed with trends and innovations in EdgeAI space.Skills:
Strong understanding of all aspects of deep learning models including model/block definitions in various frameworks, forward and backward passes, multi-GPU training and inference logic, data pre-/post-processing algorithms, weight distribution analysis and debugging
Object-oriented python product development skills for deep learning applications
Good communication, collaboration, task estimation & reporting, CI/CD via tools such as Confluence, Git, JIRA etc.Requirements:
Master or Ph.D in AI/ML or related field.
Minimum 2 years of professional PyTorch experience in training CNN models for Computer Vision applications.
At least 3 years of commercial software development experience, preferably in C++ or Python.
Knowledge of model optimization techniques, including but not limited to: Quantization, Tuning, Knowledge Distillation etc.
Knowledge of Deep Learning frameworks Like TensorFlow, TFLite, Keras and ONNX.
Good understanding of algorithms like Face Recognition, Pose Estimation, Image Segmentation, Object Detection, Gesture Recognition etc.
Familiarity with MatPlotLib, W&B and other relevant Python packages for ML applications development following is a plus.
