SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Graphcore is building the future of AI compute infrastructure. As part of the SoftBank Group, the company develops semiconductor, software, and AI technology across the complete stack—from silicon to datacenter-scale infrastructure.
This role focuses on turning large-scale system measurements into actionable engineering decisions. The team runs workloads across clusters of machines, collects detailed performance data, and determines whether systems are ready for production. The core challenge is interpreting what measurements mean and establishing confidence in system behavior.
Key responsibilities include:
- Analyzing real system performance data to answer critical questions: Is the system behaving as expected? Is performance stable enough to trust? Does it meet production criteria?
- Understanding variability, repeatability, and distinguishing signal from noise in complex systems
- Shaping measurement strategies, influencing how tests are run, and improving how results are interpreted
- Producing outputs that directly support real engineering decisions (not dashboards or reports for their own sake)
- Working across boundaries between measurement design, systems engineering, and data analysis
The role is not confined to a single discipline. You may specialize in one area but the team collectively owns all aspects of the performance analysis pipeline. The work demands careful thinking about uncertainty and evidence, preference for clarity over presentation, and willingness to challenge conclusions when data is weak.
Required qualifications:
- Strong software engineering experience across multiple projects or systems over several years
- Hands-on experience in Linux-based environments, ideally with distributed or high-performance systems
- Proficiency in Python
- Sound engineering judgment and ability to work across technical boundaries
- Experience with performance analysis, systems measurement, or related fields is valued
