Performance Benchmarking Engineer - RISC-V Workloads

Job Description
We’re looking for a hands-on engineer to support benchmarking and performance analysis of real workloads running on a custom RISC-V processor, implemented in both silicon and FPGA-based environments.

You’ll be responsible for executing, automating, and analyzing application-level benchmarks to help evaluate system throughput, scalability, and efficiency. Your work will directly influence hardware optimization and shape how we present performance data to external partners and customers.

Responsibilities:

• Run and monitor Redis, MongoDB, ElasticSearch, and other similar workloads on pre-configured systems.
• Vary workload parameters (e.g. thread/core counts, memory configurations) to explore performance boundaries.
• Capture and log key metrics like:
  • Operations per second
  • Power usage
  • Latency
  • Ops/sec per watt
• Compare results across configurations and flag anomalies or bottlenecks.
• Automate test runs and data collection where possible.
• Work with internal engineering to interpret patterns and iterate on performance hypotheses.

Must-Have:

• Running performance benchmarks (e.g., Redis, MongoDB, ElasticSearch)
• Linux system experience (Fedora 29 or similar)
• Scripting for automation (Python, Bash, or Go)
• Basic data analysis (e.g., aggregating logs, calculating throughput, visualizing metrics)
• Experience working in CI/CD environments or remote test infrastructure

Nice-to-Have:

• Redis/Mongo/Elastic performance tuning
• Experience collecting metrics like ops/sec, latency, power usage
• Familiarity with performance visualization tools (Grafana, Datadog, or similar)
• Comfort with custom or constrained environments (e.g., no package managers, legacy OS versions)

Bonus:

• Exposure to FPGA-based platforms like Zebu, FireSim, or Veloce
• Understanding of many-thread architectures or massively parallel hardware
• Prior experience benchmarking on non-x86 or pre-silicon systems
• Familiarity with profiling tools (e.g., perf, top, htop, or low-level instrumentation)
• Experience working with distributed NoSQL systems under load

About the Hardware Platform:
You’ll be benchmarking real workloads on a custom RISC-V processor, built in-house and currently deployed in both FPGA and silicon environments. The architecture includes:

• High thread count per core
• A design optimized for efficiency per watt, not peak single-threaded performance
• Unique memory behavior and thread scheduling characteristics that affect how workloads scale

You won’t need to debug RTL or low-level drivers, but understanding how workload behavior maps to threads, memory traffic, and performance bottlenecks will be key to interpreting results and guiding follow-up tests.

This is a short-term contract role with room to extend or expand scope if things go well.