Expertise
My work connects processor behaviour, software workloads and measurable performance.
Architecture and microarchitecture
Processor cores, cache and memory hierarchies, coherence, interconnects and virtual memory. Experience across IBM Power, AMD Ryzen and Qualcomm Hexagon DSP provides different perspectives on the same question: what limits useful work?
Modelling and validation
C/C++ performance modelling, pre-silicon performance verification and simulator-to-RTL correlation. I investigate execution and timing differences, validate counters, and build tools that make a model’s assumptions easier to inspect.
Workload and systems analysis
PMU/PMC analysis, workload characterisation, bandwidth and latency analysis, statistical analysis and visualisation. Tools include Linux, perf, gem5, Python and Plotly.
Research and current interests
Hardware counters and DNN workload signatures; hardware/software co-design; on-device inference and memory constraints; AI-assisted engineering workflows. Current interests are distinct from claims of production experience.