agentsclimarketplace
Sign in

Optimize llm inference

Skill hiteshbandhu/skills-i-use/skills/ai-engineer-talks/optimize-llm-inference

Runs workflows for GPU/inference talks — Groq LPU latency, OSS inference stacks, MoE networking, serverless fine-tune, frontier cluster design, Devin agent infra, Gemma sizing. Use when optimizing inference $/token, training networks, or picking open models.From its SKILL.md

Install
npx -y skills add hiteshbandhu/skills-i-use --skill optimize-llm-inference

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
  • runs commandsInstructs the agent to run 3 commands, including `cp -r skills/optimize-llm-inference ~/.claude/skills/` and 2 more.

SKILL.md

2.3 KB, 524 tokens by cl100k_base, as published. Nobody here has run it

Optimize LLM inference

Action playbook from eight AI Engineer / World's Fair talks. Do not summarize talks — pick a workflow and execute it.

Supporting files (read when needed):

  • workflows.md — workflows A–G (steps, deliverables, stop conditions)
  • source-index.md — src-NNN → talk learnings in ingest-into-skills

Optional: {SKILL_OUTPUT_DIR}/optimize-llm-inference/

Step 0 — Pick workflow

Use the decision tree below. Open the matching section in workflows.md.

What is the user trying to do?
├─ Ultra-low-latency inference (LPU/Groq-class)         → A
├─ Production OSS inference (quantization/routing)      → B
├─ MoE training network topology                        → C
├─ Fast fine-tune experiment loops                      → D
├─ Frontier cluster / SemiAnalysis planning               → E
├─ Coding agent infra (Devin-style)                     → F
└─ Pick open model tier (Gemma)                         → G

Stop summarizing once a workflow is identified — run its checklist.

Install

cp -r skills/optimize-llm-inference ~/.claude/skills/
cp -r skills/optimize-llm-inference ~/.cursor/skills/
cp -r skills/optimize-llm-inference ~/.codex/skills/

From skills-i-use or ingest-into-skills (playlists/gpus-inference-aie-world-s-fair-2024/).

Cross-cutting rules

RuleSource
Measure $/token at your concurrency[src-002 @ 10:24]
Network matters for MoE more than GPU count[src-003 @ 7:56]
Right-size model per route[src-002 @ 2:07]

Disputed steps: see source-index.md. Name workflow A–G; save artifacts to ./skill-outputs/optimize-llm-inference/ when requested; do not auto-commit.

Invocation examples

@optimize-llm-inference compare inference $/token
MoE cluster network design

What ships with it: 3 files

3.8 KB alongside SKILL.md

Gives 0 of the 12 instructions most context ai engineering skills give in 524 tokens

Counted across 1,328 of the 2,349 authors here whose files we hold, read 2026-09-06

  • Dispatch a fresh subagent for each taskin 76 of 1328, across 59 files
  • Perform spec compliance review before code quality reviewin 44 of 1328, across 34 files
  • Dispatch a final code reviewer after all tasksin 38 of 1328, across 26 files
  • Answer subagent questions before allowing implementationin 36 of 1328, across 26 files
  • Use the least powerful model capable of the taskin 33 of 1328, across 26 files
  • Create a TodoWrite list for all tasksin 32 of 1328, across 22 files
  • Perform a task review after each implementationin 31 of 1328, across 24 files
  • Extract all tasks and context from the planin 29 of 1328, across 20 files
  • Provide full task text to subagentsin 28 of 1328, across 20 files
  • Use git worktrees for isolated workspacesin 25 of 1328, across 20 files
  • Specify the model explicitly when dispatching a subagentin 23 of 1328, across 18 files
  • Execute all tasks from the plan without stoppingin 21 of 1328, across 16 files

Said here and by no other author read

  • Pick a workflow from the decision tree
  • execute the workflow checklist
  • save artifacts to the output directory
  • Measure cost per token at concurrency
  • Right-size the model per route

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.