Job Summary
Own system-level power and performance characterization and optimization across ARM-based heterogeneous SoC platforms — establishing benchmark methodology, finding bottlenecks across CPU/GPU/NPU/DSP and memory, and delivering measurable improvements to latency, throughput, thermal behaviour and battery life.
Key Responsibilities
- Define and maintain benchmarking methodology and an automated performance/power measurement lab across platform variants.
- Characterize workloads across CPU, GPU, NPU, DSP, memory and interconnect; identify and quantify bottlenecks.
- Optimize the platform stack — scheduler and CPUFreq/CPUIdle governor tuning, DVFS policy, core affinity, big.LITTLE placement, memory and cache behaviour.
- Drive power optimization: suspend/resume paths, idle-state residency, clock and power domain gating, peripheral wake analysis and low-power mode enablement.
- Measure real current consumption with power analysers and rail-level instrumentation; build power models for product use cases.
- Analyse thermal behaviour, throttling and sustained-performance envelopes; tune thermal policy.
- Profile with perf, ftrace, LTTng, GPU/NPU profilers and hardware trace; deliver actionable findings to owning teams.
- Track KPIs release over release, catch regressions in CI, and publish performance and power reports to stakeholders.
Skill Requirements
- 8–10 years in embedded systems with a strong performance and power engineering focus.
- Strong C and Python for instrumentation, automation and analysis.
- Linux performance tooling depth — perf, ftrace, eBPF, trace-cmd, systrace/perfetto.
- Linux power management internals — runtime PM, CPUIdle, CPUFreq, EAS, wakeup sources, regulator/clock frameworks.
- Hands-on power measurement with power analysers, shunt/rail instrumentation and correlated software traces.
- Ability to reason quantitatively about memory bandwidth, cache behaviour and interconnect contention.
Experience building automated benchmark suites and tracking KPIs in CI.
Other Requirements
Good-to-Have Skills
- GPU/NPU profiling and graphics or ML workload optimization.
- Thermal modelling and sustained-performance tuning.
- Standard benchmark suites — MLPerf, Geekbench-class, GFXBench, SPEC.
- Android power framework and battery-life analysis.
- Statistical analysis and dashboarding of performance data.