---
title: "Inference² — performance engineering for LLM inference"
description: "IOP Systems builds Inference², a performance-engineering suite for LLM inference: Wayfinder tunes your stack before you deploy, Sightlines watches it in production."
url: "https://iop.systems/"
---

# Inference² — performance engineering for LLM inference

Inference², from IOP Systems, helps teams measure before deployment and monitor after launch. The choice of models, runtimes, and hardware is exploding — and so is the cost of choosing wrong. Building the best inference stack in your cloud takes automation and judgment; Inference² provides both.

## Products
- [Wayfinder](https://iop.systems/wayfinder/) — Measure. Tune your stack before you deploy: run sweeps across runtimes, hardware, and configurations to find the best setup for your workload, evaluated against the theoretical performance ceiling rather than intuition.
- [Sightlines](https://iop.systems/sightlines/) — Monitor. Actionable reports from high-resolution telemetry collected across hardware, the kernel, and the runtime, analyzed to find signal in the noise and tied back to business metrics.

## Working with
Intel, Roblox, Fastly, eBay, and Momento.

## Elsewhere on this site
- [Company & team](https://iop.systems/company/) — the operators behind IOP Systems.
- [Open source](https://iop.systems/oss/) — Rust crates and tools including Rezolus, rpc-perf, and llm-perf.
- [Blog](https://iop.systems/blog/llm-inference/) — a systems view of inference and the serving ecosystem.
- [llms.txt](https://iop.systems/llms.txt) — a concise, machine-readable map of this site.