Full job description
Rivos Accelerator Performance team is seeking highly motivated candidates to perform post-silicon performance validation activities for Rivos accelerator design. The primary role is to engage in performance validation and analysis of typical workloads on server compute silicon.
Key Responsibilities
As a Post-Silicon Performance Engineer, you will own or participate in the following:
* Analyze workloads and their performance to drive the performance analysis, optimization and correlation efforts on Rivos' accelerator design.
* Quantify and analyze performance of typical server workloads (data analytics, graph algorithms, and AI/ML), and correlate measurements with simulation results
* Collaborate cross-functionally with design, architecture, systems, and software teams to enable enterprise customer use-case performance measurements
* Perform silicon performance measurements and correlate with partition simulations and full-chip projections
* Improve use-case power/perf ratio through tuning of hardware and software settings
* System level performance debugging and silicon performance per watt tuning
* Automate performance validation infrastructure
* Develop and execute silicon characterization tests for the Rivos accelerator.
Minimum Qualifications
* 5+ years of hands-on experience in bring-up and validation of complex SoCs.
* Experience scripting in Python or other programming language.
* Experience in object-oriented programming with C and C++
* Experience in bringing up silicons, performing functional and performance validation, and debugging failures.
* Familiar with the LLM/ML workloads
Preferred Qualifications
* Knowledge and experience with common performance benchmarks such as MLPerf
* Experience working multi-functionally with people of very diverse backgrounds
* Demonstrated communication, influence, and negotiation skills
* Technical documentation skills
* Initiative in taking ownership & responsibility to drive deliverables
* Able to think outside the box and come with creative solutions for lab validation.
