Linux System Programming Performance Engineer

Role/Skill : Linux System Programming Performance Engineer

Experience : 8+ Years

Location : Bangalore

Job Description :

Linux System Software Performance Engineer responsible for performance measurement, trace collection, and triage across bare metal and virtualized platforms, with focus on hard real time behavior and system level performance derates.

Core Responsibilities PAVS KPIs & Deliverables :

- Own end to end performance measurements across Linux system software stacks

- Collect traces, triage issues, and perform root cause analysis for performance deviations

- Identify, quantify, and report system bottlenecks impacting PAVS KPIs

- Generate trace backed evidence for performance sign off and regressions

- Collect and analyze traces for : hard real time latency, Interrupt latency, scheduler latency, wake up latency

- Jitter and determinism characterization using kernel traces

- Triage real time violations and correlate with kernel / driver paths

- Hypervisor vs Bare Metal Performance Derates

- Measure and quantify performance derates in virtualized vs bare metal environments

- Perform subsystem level triage and trace analysis for :

1. CPU (scheduling, vCPU overhead)

2. GPU (submission latency, virtualization overhead)

3. NPU (compute and scheduling impacts, if applicable)

4. Memory (bandwidth, latency, contention)

5. Disk / Storage (I/O latency, throughput)

6. Network (latency, throughput, virtualization overhead)

- Identify virtualization overheads and provide optimization insights

- Performance Tooling & Trace Collection

- Kernel and user space profiling tools : perf, ftrace, trace cmd

- Scheduler, IRQ, I/O trace collection

- Performance baselining and regression tracking

Required Skills :

- Strong Linux system programming experience

- Must have strong hands on programming in C, C++, or Python (any one)

- Linux device driver development experience

Good to Have :

- Kernel, driver, or middleware performance exposure

- Experience with real time Linux or hypervisors

- Not Mandatory but good to have exposure

- PnP (Power and Performance) experience