Traversal raises $120 million for AI site reliability engineering
Traversal has raised $120 million at a $1 billion valuation to expand its AI site reliability engineering platform, which helps engineering teams identify the root causes of failures in live enterprise software. Site reliability engineering, or SRE, is the work of keeping digital services available and restoring them when something goes wrong. Traversal brings AI into that operational work, examining evidence from running systems to identify the causes of failures and help engineers respond.
Announced on 3 November 2026, its Series B round was co-led by venture capital firms Kleiner Perkins, Redpoint Ventures and Khosla Ventures, with participation from Sequoia Capital, Battery Ventures and Amex Ventures. It brings Traversal’s total funding to $173 million.
Why a software failure can be difficult to trace
A payment that fails on a website may depend on a chain of services spanning identity checks, inventory, databases and external payment processing. Trouble in one component can spread through that chain, producing errors far from its origin.
Engineers investigating the incident must distinguish those downstream symptoms from the initiating fault. They examine logs recording system activity, metrics measuring performance and traces showing how requests travel between services, alongside recent code and configuration changes.
Consider a database slowdown that causes several applications to time out. Alarms may arrive together even though the fault originates in one shared component. Finding that starting point is a problem of cause and effect, making it a natural application for the research backgrounds of Traversal’s founders.
Traversal’s founders and causal machine learning
Chief executive Anish Agarwal and co-founder Raaz Dwivedi bring research in causal machine learning: methods concerned with understanding cause and effect. Agarwal holds a doctorate from the Massachusetts Institute of Technology and a faculty appointment at Columbia University; Dwivedi earned his doctorate at the University of California, Berkeley, and is a professor at Cornell Tech.
Fellow co-founder Raj Agrawal, also an MIT doctorate holder, worked on graphical models and how changes in gene regulatory networks cause disease; this is research with parallels to tracing failures through connected software services. Ahmed Lone brought engineering and operational experience from quantitative trading firm Citadel Securities.
How Traversal’s Production World Model works
Traversal’s architecture starts with what it calls its Production World Model: a continuously updated representation of a customer’s live software environment.
The model brings together services, their dependencies, deployments, configurations, code and telemetry, in other words the operational data generated by running systems.
This addresses a practical problem with investigating failures: an architecture diagram or written description can become outdated as teams deploy changes. An investigation needs the relationships and conditions that existed when the incident occurred.
In the database example, this means establishing which applications use the database and how they connect to it. Traversal’s Knowledge Bank adds operational context, such as previous incidents and troubleshooting guidance, that helps its agents interpret those relationships.
With those connections and records assembled, Traversal’s Causal Search Engine can investigate possible explanations.
Finding root causes with the Causal Search Engine
According to Traversal, the Causal Search Engine evaluates thousands of hypotheses in parallel, checking them against the system’s dependency structure, timing and patterns of failure propagation.
Using the connections mapped by the Production World Model, the engine can examine whether the database slowdown preceded errors in the applications that depend on it. It can also investigate competing explanations, such as a configuration change that affected both the database and those applications.
Traversal says its engine eliminates explanations that conflict with those constraints and returns a diagnosis supported by evidence, together with a remediation path. This gives engineers an explanation they can examine alongside the proposed corrective action.
Enterprise customers and incident response
Traversal’s customers include Brazilian digital bank Nubank, legal AI company Harvey, cloud infrastructure provider DigitalOcean, food and beverage group PepsiCo and financial services company American Express.
The company reports more than 150,000 investigations last quarter, supporting over 4,000 engineers. The investigation count covers broader operational enquiries as well as incidents. Separately, Traversal says roughly two-thirds of product usage now extends beyond root cause analysis, including assessing alerts, examining release risks and answering production support questions.
A case study published by Traversal describes how managed cloud hosting provider Cloudways integrated its technology into Cloudways Copilot. Customers can investigate problems and apply recommended remedies through a “SmartFix” button, reducing the exchanges previously needed with support engineers.
In one reported incident, a WordPress website slowed down, processor usage surged and visitors encountered errors. Traversal began investigating at 2:08pm and identified abusive IP addresses—the network addresses of traffic sources—overwhelming the site by 2:10pm. Two minutes later, it proposed blocking those addresses and restarting affected services; the case study records resolution following a single click at 2:13pm.
In this deployment, a user initiates the recommended fix through the support interface, connecting diagnosis with corrective action.
Traversal Workers and the link to AI coding
Alongside integrations such as Cloudways Copilot, Traversal offers agents that work in the channels engineers use to coordinate their response. Its Incident Workers are generally available and investigate incidents in workplace messaging platforms Slack and Microsoft Teams.
Rather than requiring an engineer to initiate every query, these agents decide when to engage and contribute findings as the incident develops. A Workers console provides a place to manage agents and start shared sessions through the web.
Alert Workers, available in public beta, extend that approach to incoming alerts. They group repeated signals, assess unusual behaviour and surface alerts with supporting reasoning, helping teams decide which events need attention.
Traversal also offers a Model Context Protocol integration, using a standard that connects AI applications to external tools and information. This lets coding agents draw on its understanding of production while developers work on software.
The new funding will support hiring across research, engineering and commercial roles as Traversal develops this connection between software creation and operation. The practical benefit is continuity: information gathered while investigating a running service can also inform the engineers and coding agents preparing its next update.
Liked this article? You can support our independent journalism via our page on Buy Me a Coffee. It helps keep MoveTheNeedle.news focused on depth, not clicks.