Computer Vision and NN/ISP Algorithm Toolchain and Optimization Engineer
AmbarellaJob Title
Computer Vision and NN/ISP Algorithm Toolchain and Optimization Engineer
Role Summary
Design and develop optimization toolchains for neural networks, transformers, LLMs, and ISP models to improve efficiency and performance on Ambarella platforms. Work within the AI/vision algorithms team to enable hardware-aware model optimization, porting, and deployment.
Experience Level
Mid-level. Years of experience not specified.
Responsibilities
Primary responsibilities focus on building and applying model optimization tooling and supporting deployment on edge vision processors.
- Develop optimization toolchains for computer vision and ISP models (denoising, super-resolution, detail enhancement).
- Perform hardware-aware model optimization and port models to Ambarella platforms.
- Research and evaluate pruning, quantization, fine-tuning, and other optimization techniques for CNNs, transformers, and LLMs.
- Provide technical support and solutions to customers for model optimization and deployment.
Requirements
Core skills and desirable experience for the role.
- Must-have: Background in machine-learning-based computer vision or ISP algorithms.
- Must-have: Proficient in Python and C/C++ programming.
- Must-have: Familiarity with ML frameworks such as PyTorch, TensorFlow, or Hugging Face.
- Nice-to-have: Experience with model optimization (pruning, quantization) and model deployment.
- Nice-to-have: Experience fine-tuning ISP models and porting models to embedded/hardware platforms.
Education Requirements
Not specified.
About the Company
Company: Ambarella
Headquarters: Santa Clara, California, USA
Ambarella is a leader in computer vision and video processing technology, providing advanced solutions that enhance the performance of video applications. Focused on quality and innovation, Ambarella develops products for diverse uses, including autonomous vehicles, surveillance systems, and smart cameras, aiming to deliver pristine imagery and efficient compression while minimizing power consumption.
