Notes on systems
that do the work.
Technical writing on systems, AI infrastructure, and the quiet details that make software reliable.
01 / published note
Latest note
Edo Tensei: Checkpoint and Restore Live AI Workloads
How I built a system that freezes running AI workloads—from Python processes to CUDA and vLLM—and brings them back ready to serve.
AI RuntimeGPU SystemsLinuxInferenceOpen Source