Inference Optimizer
$2.99OfficialOptimize model inference: quantization, distillation, batching, speculative decoding, and serving infrastructure tuning.
datainferencequantizationllm-servingperformance· by SkillingMain
What you get
- ✓9-step procedure
- ✓6 pitfalls to avoid
- ✓Installs into 6 tools
- Version
- v1 →
- Last updated
- today
- Length
- 3 min read
- Requires
- Best with a strong model (Claude Sonnet 4)
Works in: Claude Code, Codex, Cline, opencode, OpenClaw, Hermes · Handles multi-file projects
Preview
When to use
Use this skill when inference cost or latency is the bottleneck — for LLM serving, CV pipelines, or high-QPS ML APIs. It covers quantization (PTQ, QAT), distillation, dynamic and continuous batching, speculative decoding, KV-cache management, and serving-stack tuning (vLLM, TGI, TensorRT-LLM, Triton). Reach for it whenever tokens/sec or $/token is blocking product or margin.
Inputs to gather
- Model type and size (LLM, CNN, transformer encoder, classical ML)
- Serving latency SLA (TTFT, per-token, end-to-end) and throughput target
- Hardware (GPU type/count, CPU-only, edge)
- Current cost per request and budget ceiling
- Quality bar: how much accuracy degradation is acce
…
🔒 Buy once ($2.99) to unlock the full playbook, download it, and install it in every tool you use.