Rivos is developing optimized Deep Learning operators for its SIMT (Single Instruction Multiple Threads) machine, providing optimal use of the compute units the HW exposes. You will contribute to development and optimization of many operators used for both training and inference of Deep Neural Networks. In this process you will be able to influence the architectural decision of the HW engine to deliver more performant and more power efficient solutions. In a vertical development approach you will be contributing extensively to all the other parts of the solution: client software, compiler, runtime, simulator to help define the next generations of our solution.
Responsibilities
As a Deep Learning Libraries engineer, you will own or participate in the following
design and implement critical parts of the DL operators libraries, including kernels used by PyTorch
contribute to the performance analysis flow to guide optimization work
contribute to the functional and performance ISA simulators
collaborate cross-functionally with Silicon design, architecture experts, and other teams across the company
Requirements
at least 3 years of experience numerical library development
strong C++ programming skills
strong knowledge of parallel programming languages
experience with PyTorch a plus
strong background in dense linear algebra software
excellent skills in problem solving, written and verbal communication, excellent organization skills, and highly self-motivated.
ability to work well in a team and be productive under aggressive schedules
Education and Experience
PhD, Master’s Degree or Bachelor’s Degree in technical subject area.
