Partnering with Empirik: Building the Autonomous Infrastructure Engineer
Sequoia Bogomil Balkansky ● Covered by 2 sources
Sequoia backed Empirik with $21M to track infrastructure changes, not just outages. It wants to stop bad deploys before they hit production, which is a lot less glamorous and a lot more useful.
Based on reporting by Sequoia, Bogomil Balkansky — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Sequoia has spun out a new company called Empirik with $21 million in funding, and the pitch is refreshingly unsexy: track the changes that break infrastructure before they break it. The company started inside Sequoia, built by Avon Puri and Sudheer Dhurjati, who both spent years watching enterprise IT stay stuck in reactive mode. Puri had run infrastructure at Rubrik and VMware, where the same problem kept showing up in different clothes. Logs and alerts were telling people that something had already gone wrong. They were not telling them why.
Empirik’s bet is that the real story lives in change. A new identity policy, a Terraform provider update, a feature flag flipping in production — those are the moments that matter. The company argues that enterprises have treated that information as paperwork or folklore, scattered across tickets, email, Slack, and compliance records. That may be enough to satisfy an auditor. It is not enough to run infrastructure well. Empirik wants to turn those changes into live context, so the system can understand intent, assess risk, and decide when to act alone and when to pull in a human.
The product is being pitched as an “Autonomous Infra Engineer,” but under the hood the idea is more old-school than the branding suggests. Empirik is building what it calls an infrastructure compiler: a unified operational graph stitched together from cloud systems, identity providers, CI/CD, SaaS tools, and more. The company says that graph maps thousands of relationships across assets and accounts, and lets it answer basic but brutal questions in real time: what changed, who changed it, and whether that change caused the incident. It is already wired into agentic workflows through MCP and can connect to tools like Claude and ChatGPT.
And this is not a greenfield toy. Sequoia says the team built for Fortune 500 reality from the start, meaning messy hybrid stacks, on-prem systems, identity layers, and the usual pile of legacy software nobody gets to delete. The traction claims are pointed too. TCBPay, which processes more than $1.5 billion in annual volume, says it cut the time to catch sensitive config changes from 30 minutes to 10 seconds. Avahi is using Empirik inside GitHub Actions to evaluate and block risky changes at the source. Guardant Health expanded from production into all dev environments and on-prem, and a global Fortune-50 CPG company is widening its use as well.
The broader bet is simple: observability found the smoke, but Empirik wants to find the match. That’s a better business and a better product, because nobody gets paged for a theory. They get paged for a change that just landed in production.
My take — AI-written commentary, not fact-checked reporting
This is the rare infra pitch that sounds useful instead of merely expensive. The industry has spent years worshipping alerts, then acting shocked when alerts arrive after the damage. A system that watches the change itself is just common sense with a seat at the table, which is probably why it took so long to become fashionable.
Read more about this at: Sequoia