Amayes Djermoune

• Engineered an automated I/O energy benchmarking testbed in C, C++ and Python on Grid'5000 (large-scale French HPC Infrastructure), reducing manual setup overhead by 56% and enabling reproducible power profiling across 2+ energy consumption protocols.

• Diagnosed and resolved a systematic bug in the shared energy interpolation formula affecting all I/O protocols, restoring accurate per-operation energy measurements and reducing latency deviation to <0.1ms post-fix.

• Reduced I/O latency measurement overhead by 60% by diagnosing two root causes in the I/O stack bypass protocol, correcting lseek timing order, and enforcing per-request cache flushing (drop caches).

• Architected a geospatial data platform covering 300+ French inter-municipal authorities (EPCIs) using PostgreSQL/PostGIS QGIS, enabling local governments to visualize and broadband coverage gaps.