Pomegra
Sequoia spinout Empirik raises $21M to predict IT infrastructure outages by mapping system change ripple effects before they cascade into failures.
- Empirik's $21M seed was led by Sequoia Capital, with S32, Canapi Ventures, and Alumni Ventures participating.
- The company was incubated inside Sequoia since 2023, co-created by two Sequoia IT leaders, before launching independently on September 1, 2026.
- Early production customers include S&P Global, Guardant Health, and multiple unnamed Fortune 500 companies.
Lead
Empirik launched as an independent company on September 1, 2026, closing a $21 million seed round led by Sequoia Capital with participation from S32, Canapi Ventures, and Alumni Ventures. The San Francisco-based startup had been incubated inside Sequoia since 2023, built around a core premise that most IT outages are predictable - if you model what a change will do before it runs, not after it breaks something.
What Does Empirik Actually Build?
The platform centers on what Empirik calls an infrastructure compiler: a continuously updated operational model spanning on-premises systems, public cloud environments, and SaaS applications. It treats virtual infrastructure components - IAM policies, access control lists, routing tables - as first-class objects, mapping the dependency graph across the entire stack.
When an infrastructure change is queued, the system calculates its potential blast radius by tracing cascade paths through that graph. Low-risk changes are permitted to execute autonomously. Higher-risk changes receive automated guardrails. The highest-risk updates are flagged for human review before anything touches production. The company describes the platform as an autonomous traffic cop for infrastructure changes, though the analogy undersells the technical complexity: the system is doing probabilistic failure modeling at the graph level, not just comparing change logs.
Why Is Predicting Outages Harder Than It Sounds?
Modern enterprise infrastructure is a dependency graph with thousands of nodes and no clean documentation of how they interact. A configuration change to an IAM policy might cascade into a broken API call that surfaces three hops away, hours later, in a service that nobody connected to the original ticket. Traditional monitoring catches failures after they propagate. Change management processes catch obvious risks but rely on human review that doesn't scale to the volume of changes large infrastructure teams make daily.
The result is a persistent gap: organizations know their systems are fragile at the points of change, but they have no systematic way to evaluate risk before committing. Empirik's bet is that machine-speed graph modeling can close that gap in a way that neither human review nor reactive alerting can.
Who Built This and Why Does Sequoia's Role Matter?
Kartik Chandrayana, who previously ran more than 30 infrastructure products at Salesforce, leads Empirik as CEO. The company was co-created by Avon Puri and Sudheer Dhurjati, both Sequoia IT leaders, with backing from Sequoia partner Bogomil Balkansky. The incubation structure is notable: Sequoia didn't write a check into an external founder's idea - it built the company internally and spun it out. That means Empirik's initial customer relationships, product assumptions, and network came directly from one of the most connected firms in enterprise software. The seed round formalizes the independence while Sequoia retains its position as lead investor.
Sequoia's internal incubation model has produced a handful of companies over the years, but it remains uncommon. The structure gives portfolio companies unusual early access to enterprise CIOs and infrastructure leaders, which explains how Empirik reached production deployments at S&P Global and Guardant Health before its public launch.
What Comes Next for Infrastructure AI?
The infrastructure observability market has attracted significant investment over the past three years as cloud complexity has outpaced human operational capacity. Where earlier entrants focused on log aggregation and anomaly detection - reactive approaches that identify problems as they emerge - a newer cohort is attempting preemptive modeling. Empirik sits firmly in that second wave.
The $21 million seed is large for the stage, which signals that investors expect a significant infrastructure buildout before the company reaches repeatability at scale. Graph modeling across heterogeneous infrastructure stacks is computationally expensive, and enterprise sales cycles for systems that sit in the critical path of production changes are long. The capital also gives Empirik runway to absorb the integration complexity that comes with Fortune 500 deployments, where environments rarely resemble the clean configurations that AI systems train on.
Valuation was not disclosed.
Outlook
Empirik enters the market with a structurally unusual advantage: three years of internal development inside Sequoia, production deployments already underway, and a $21 million seed giving it the runway to close longer enterprise sales cycles. The core technical challenge - building a dependency model comprehensive and accurate enough to be trusted for autonomous change execution - remains unsolved at scale. Early customer traction suggests the approach works in controlled environments. The harder test is whether it holds across the full entropy of large enterprise infrastructure, where undocumented dependencies and legacy systems routinely defeat cleaner models.
