Senior Principal DSP Engineer, Machine Learning – A004286
Location: Campbell, CA, USA
Job Type: Full Time Employee
Career Level: Experienced
Primary Responsibilities
Development of Fixed Point Models of Deep Neural Networks (DNN) using MATLAB and C/C++. Perform fixed point training and inferencing; compare accuracy of fixed point training and inferencing against their floating-point counterparts. Explore and develop techniques for accelerating training for a variety of fixed point DNN models, while maintaining numerical stability that lead to convergence. Benchmark performance and compare against other platforms from the competitors. Understand and master the tradeoffs between computation and storage requirements as it relates to the requirements for shrinking model sizes for Inference type applications. Explore and implement methods for efficient Transfer Learning.
Required Skills and Experiences
- Develop and test MATLAB and C/C++ programs as a member of the development team working on the fixed point models of DNNs. Compare/benchmark the results against their floating-point counterparts. Participate in the applied research and/or advanced development activities in Machine Learning Algorithms. Must be a team player.
- Exposure to Machine Learning; Exposure to Training and Inference processing for Convolutional Neural Networks (CNN); Proficiency with MATLAB is a must. Must have experience with fixed point implementations for Digital Signal/Image Processing or Digital Communications. Knowledge of Adaptive Signal Processing. Conference and/or Journal Publication(s) in at least one of following areas: (a) Digital Signal/Image Processing, (b) Adaptive Signal Processing using LMS (Stochastic Gradient Descent) or its variants.
- Minimum education level required: Ph.D or MS in Electrical Engineering.
- Minimum years of experience required: Ph.D. plus 5 years or MS plus 10 years.
Desired Skills
- Proficiency with C and/or C++ is preferred
- Conference and/or Journal Publication(s) related to Machine Learning is a plus
- Exposure to Keras, Tensorflow, Caffe etc. is desired