Self Hosted Inference Server Ops
by Lawrence Bai
Keep self-hosted LLM servers alive: hot-updates, armed restarts, swap recovery, and mystery triage.
by Lawrence Bai
Keep self-hosted LLM servers alive: hot-updates, armed restarts, swap recovery, and mystery triage.
by Lawrence Bai
Selection, RAM-sizing, and hardware decision tree for large local MoE/dense LLM inference.
by Lawrence Bai
The only marketplace playbook for running NVFP4-quantized large MoE models on a single consumer RTX box: architecture decision tree, tuned parameter bands, real measured tok/s and VRAM numbers (all tagged by evidence class), and 12 field-tested failure modes with fixes.
by Lawrence Bai
Get paid, not ghosted: statute-cited contract clauses, acceptance/invoice templates, tax-withholding quick-ref, and the 6-rung legal collection ladder for Taiwan freelancers. The paid core of the win-the-bid → get-paid system.