Empirik Spins Out of Sequoia With $21M to Forecast Infrastructure Outages

Updated September 1, 2026 7:54 pm

Empirik has launched as an independent company with $21 million in seed funding to help engineering teams predict outages before they happen. The startup, incubated by Sequoia Capital, is betting that AI can move infrastructure operations from reactive firefighting to proactive prevention.

Founded around the idea that software systems can be monitored for risky changes and failure patterns before incidents occur, Empirik is positioning itself as an autonomous layer for DevOps and site reliability teams. Its backers include Sequoia, Canapi, and Alumni Ventures, and the company has already signed customers ranging from startups to large enterprises, including S&P Global and Guardant Health.

Why Empirik matters now

Software teams are shipping code faster than ever, but the infrastructure underneath those applications is also becoming more complex. That combination has made outages more expensive, more frequent, and harder to diagnose quickly.

Empirik is designed to address that gap. Instead of waiting for a service to fail and then tracing the damage after the fact, the company’s platform watches for configuration and systems changes, then infers how those changes could spread across an environment. The goal is to warn operators early, block the riskiest updates, and reduce the burden on engineers who spend too much time on repetitive incident response.

The company emerged from work inside Sequoia, where longtime infrastructure operator Avon Puri and Sequoia colleague Sudheer Dhurjati first identified the opportunity. The pair believed large language models had matured enough to help systems teams do more than simply summarize logs or speed up troubleshooting. In their view, AI could participate in decision-making about infrastructure risk.

How did Empirik get started?

Empirik began as an internal incubation project at Sequoia in 2023 and was later shaped into a separate business. The startup added a familiar enterprise software operator to the founding story earlier this year when Kartik Chandrayana, formerly chief product officer at Quantum Metric and vice president of observability at Salesforce, became chief executive.

The company’s roots matter because observability tools already fill data pipelines, dashboards, and alerting systems across modern IT stacks. Empirik is trying to go a step further by understanding dependency chains and change impact, rather than simply reporting that something has already gone wrong.

Sequoia partner Bogomil Balkansky described the company as targeting a large, persistent spending category in enterprise infrastructure, while arguing that many existing tools still struggle to understand how different systems depend on one another. He said Empirik is meant to act like an autonomous traffic controller, allowing routine changes through, applying guardrails to larger updates, and escalating dangerous ones for human review.

That framing helps explain why the firm believes the product has room to carve out a distinct market position. It is not just another dashboard or alerting layer. It is intended to sit closer to the operational decision-making process.

What the product does

Empirik’s platform tracks changes across infrastructure and attempts to predict the ripple effects those changes could create. In practical terms, that means the software looks for the kinds of updates that might be harmless, the ones that deserve caution, and the rare but dangerous ones that should be examined by people before they are deployed.

The startup describes the system as a kind of autonomous infrastructure engineer. For teams that are stretched thin, that could mean less time spent on routine triage and more time on strategic reliability work, architecture decisions, and higher-value engineering projects.

Empirik’s founders are openly drawing a parallel to the rise of coding assistants. Just as Cursor and Claude Code have helped developers write, refactor, and review software faster, Empirik wants to help infrastructure teams keep pace with the speed of modern software delivery.

What makes it different from observability tools?

Empirik differs from traditional observability products because it is built to reason about change, not just detect symptoms. Many tools can tell operators that latency increased, a service is failing, or error rates are rising. Empirik aims to infer whether a proposed change is likely to trigger trouble before the problem appears in production.

That distinction could prove important in environments where teams use many services, deployment pipelines, cloud resources, and third-party systems. In those settings, the hardest part is often not collecting data but understanding how one action cascades into another across the stack.

Who is using Empirik so far?

Empirik says it has already secured customers across a range of company sizes since its public launch earlier this year. Those users include startups and several Fortune 500 organizations, among them S&P Global, Guardant Health, and a major consumer packaged goods company that the startup did not name.

Early customer adoption suggests there is interest in tools that can reduce downtime and help operators make faster decisions with more confidence. It also signals that the company is targeting an audience accustomed to buying enterprise software that can justify its cost by preventing expensive outages.

Why investors see a market opportunity

Enterprise infrastructure is one of the most durable spending categories in technology. Companies need monitoring, incident management, logging, configuration control, and alerting regardless of whether they are in a startup phase or operating at global scale. As systems become more distributed, the pressure to prevent outages rather than merely respond to them keeps rising.

That is the bet Sequoia and Empirik’s other investors are making. If the product can reliably anticipate outages and reduce incident volume, it could save customers money in several ways at once: less downtime, fewer engineering hours lost to troubleshooting, and lower operational risk.

There is also a broader strategic reason the opportunity looks timely. AI has already changed how code is written. The next frontier may be how software operations are managed. Companies that can automate more of the infrastructure lifecycle may gain an efficiency edge as development cycles accelerate.

How does agentic AI fit in?

Agentic AI is relevant because it moves systems from passive analysis toward action. In Empirik’s case, that means software that does more than recommend; it evaluates, decides, and in some situations acts like an automated gatekeeper for infrastructure changes.

Chandrayana has argued that as AI speeds up application development, the teams responsible for reliability need tools that can keep pace. The company’s thesis is that infrastructure engineering will need its own generation of AI-native workflows, just as software development did.

How competitive is the space?

Empirik is entering a crowded but still evolving market around site reliability, observability, and AI-assisted operations. Some startups and incumbents are already using AI to help identify problems, summarize alerts, or guide remediation. The challenge for any newcomer is to stand out by solving a concrete pain point in a measurable way.

Sequoia’s Balkansky says Empirik currently occupies a category of its own, though he views it as complementary to AI SRE offerings such as Resolve and Traversal, the latter also backed by Sequoia. That suggests the market may not be a winner-take-all race, but rather a set of overlapping products that attack different layers of the operations stack.

Still, the company’s success will depend on execution. To win enterprise buyers, it will need to prove that its predictions are accurate, its recommendations are trustworthy, and its automation can be adopted without creating new operational risk.

Why the Sequoia connection matters

Sequoia’s involvement gives Empirik instant credibility in a market where trust is essential. Infrastructure software often handles mission-critical systems, so buyers want confidence that a vendor understands their environment and has the technical depth to avoid introducing new failure modes.

That background also helps explain why the startup was incubated rather than launched cold. Sequoia had internal expertise in infrastructure, operational workflows, and enterprise buying behavior, which likely shaped the product from the beginning. The company’s origin story is less about a consumer-style startup gamble and more about a venture firm identifying a pain point from inside the machinery of its own operations.

For Puri, the move from running internal systems at Rubrik and VMware to helping build a new company reflects a familiar arc in enterprise software: operators who have lived the problem often become the best founders for solving it.

What happens next?

Empirik now has the capital and independence to expand beyond incubation and prove whether its model can scale. The company will need to deepen its integrations, improve the accuracy of its risk predictions, and show that its automation can function safely in complex production environments.

If it succeeds, the startup could become part of a broader shift in enterprise operations, where AI is not only generating code but also helping determine whether a change should ever reach production. That is a meaningful evolution for a field built around preventing expensive mistakes.

For now, Empirik’s launch signals that investors and operators alike are looking for AI products that do more than chat or summarize. The next valuable layer may be the software that keeps the rest of the software running.

Bottom line

Empirik is trying to turn infrastructure reliability into a predictive AI problem. Backed by Sequoia and launched with $21 million, the startup wants to help enterprises catch dangerous system changes before they trigger outages.

That mission puts it at the intersection of AI, enterprise software, and site reliability engineering — and in one of the most important unsolved areas of modern tech operations.