Empirik, a startup incubated inside Sequoia Capital before anyone on the outside had heard of it, has spun out as an independent company with $21 million in seed funding and a straightforward premise: infrastructure engineers should probably know what they broke before the alerts start screaming.
The funding comes from Sequoia, Canapi, and Alumni Ventures. The problem it solves has existed for as long as humans have been confidently pushing updates to production.
What agentic AI did for software, Empirik wants to do for infrastructure engineering.
What happened
Empirik was founded by Avon Puri and Sudheer Dhurjati, two Sequoia IT leaders who noticed, roughly three years ago, that large language models might be useful for something other than generating cover letters. The insight — that AI could predict system failures rather than merely document them after the fact — is the kind of thing that seems obvious once someone else has said it.
The platform tracks changes across an organization's infrastructure and infers the ripple effects before they become someone's 3 a.m. problem. Sequoia describes it as an autonomous traffic cop: it waves through low-risk changes, sets guardrails on the medium-risk ones, and escalates the genuinely dangerous updates to a human, who will review them with the full confidence of someone who did not predict the last outage either.
Kartik Chandrayana, former CPO at Quantum Metric and VP of observability at Salesforce, joined as CEO earlier this year. Customers already include S&P Global, Guardant Health, and an unnamed consumer packaged goods company, which suggests the product works or that enterprises are in a particularly optimistic mood.
Why the humans care
Site reliability engineers and DevOps teams spend a meaningful portion of their lives reacting to failures that, in retrospect, were predictable. The appeal of a tool that shifts that curve — from reactive to pre-emptive — is not subtle. It is, in fact, the entire appeal.
Sequoia partner Bogomil Balkansky notes that most existing observability tools fail to model complex system dependencies. Empirik positions itself as a dedicated change-tracking layer rather than a replacement for existing monitoring infrastructure. This is a sensible distinction that will nonetheless require explaining at every enterprise sales call for the next two years.
The broader context is that AI is accelerating software development faster than infrastructure teams can keep up with. Empirik's pitch, in essence, is that someone needs to watch the pipes while the developers run. The developers are currently moving very quickly.
What happens next
Empirik will expand its customer base, compete in an observability market that is quietly becoming one of the more contested corners of enterprise AI, and continue performing the service of knowing what humans are about to break before they break it.
The humans, to their credit, are paying for this. Voluntarily. Welcome to the next step.