Traversal raises $120M at $1B valuation to build AI-powered site reliability engineering for Fortune 500

Traversal raises $120M at $1B valuation

Traversal, founded in January 2024 by Columbia professor Anish Agarwal, has raised $120M at a $1B valuation with Kleiner Perkins as lead investor. The company builds an AI site reliability engineer that uses causal machine learning to identify root causes of system incidents at Fortune 500 companies including American Express and PepsiCo.

The core problem

When a system goes down at a large enterprise, everything breaks at once. Thousands of engineers, each with limited context, are staring at fragmented tooling and telemetry. Finding the root cause, Agarwal says, is like finding a needle in a haystack full of fake needles. His research background in causal machine learning, specifically the study of cause-and-effect relationships in data, is what Traversal applies to that problem.

“We started the company in January 2024... our research is in a very specific niche field called causal machine learning... finding root cause is like finding a needle in a haystack with fake needles everywhere... We raised $120,000,000 and a million dollar [valuation]... probably the biggest users of the product now are becoming the coding agents themselves.”

The difficult early stretch

The company spent roughly the first 18 months working through what Agarwal describes as a desert period. They were pulled toward large enterprises early, which meant dealing with production-scale complexity before the product was ready for it. The breakthrough came about six months ago. Since then, usage has been scaling with existing Fortune 500 clients rather than expanding down-market.

Why CIOs are on edge

The fear driving demand isn't just reliability, it's ambiguity. When something breaks, security teams and engineering teams can't tell whether it's a traditional outage, a rogue internal agent over-stressing a database, or an active breach. The symptoms, users can't log in, the site slows down, look identical. At the same time, engineers are writing less of the code themselves, which means institutional context about what should be running and how is declining. Every incident feels like a red alert, Agarwal says, and patching security vulnerabilities without thinking through reliability implications is itself a major source of new incidents.

Usage-based pricing and an unexpected growth driver

Traversal prices on consumption: clients pay per investigation triggered. The product originally focused on live incidents, but enterprises are now using it earlier in the development cycle, flagging whether code is likely to break in production before it ships. The fastest-growing user, perhaps unexpectedly, is coding agents themselves. They call Traversal to check whether a proposed deployment will cause a reliability or security issue before executing it. That agent-to-agent workflow is now what Agarwal describes as a first-class use case.

The ambition is to be the Cursor equivalent for site reliability engineering — a flexible platform woven into the development loop rather than a point solution waiting for something to go wrong.