Sequoia's Spin-Out Empirik Raises $21M to Catch Server Outages Before They Happen
Two IT veterans who spent years firefighting tech failures at Rubrik and VMware built an AI tool that spots trouble before the system crashes. Now it's a real company with serious backing.
Key points
- Empirik raised $21 million in seed funding from Sequoia, Canapi, and Alumni Ventures, announced Tuesday.
- The company was incubated inside Sequoia Capital in 2023 before spinning out as an independent business in 2025.
- Customers already include S&P Global and cancer-testing firm Guardant Health.
- Former Quantum Metric CPO Kartik Chandrayana joined as CEO earlier this year.
Most tech outages follow the same painful script: something changes, something breaks, engineers scramble at 2 a.m. Empirik wants to flip that script entirely.
The startup, first reported by TechCrunch AI, grew out of a very specific frustration. Avon Puri spent more than a decade running the computers and networks that keep large companies alive, first at data-management firm Rubrik, then at cloud software company VMware. When he joined investment firm Sequoia Capital as its chief digital and information officer in 2020, the problem followed him.
What does Empirik actually do?
Simply put, it watches every change made to a company's tech systems and works out what could go wrong before anything does. Think of it as a highly experienced engineer who never sleeps and reads every update ticket the moment it comes in.
Puri and Sequoia IT colleague Sudheer Dhurjati spotted that large language models, the AI technology behind chatbots like ChatGPT and Claude, had become capable enough to reason about cause and effect inside complex systems. Around three years ago they started building a tool to do exactly that.
The result is Empirik. Feed it your infrastructure, and it tracks changes across the whole environment, mapping how one tweak to one server might ripple into a failure three steps downstream. Sequoia partner Bogomil Balkansky describes it as an autonomous traffic cop: it waves low-risk changes through, places guardrails on medium-risk ones, and flags the most dangerous updates for a human to approve.
Who is this actually for?
The tool targets DevOps and site reliability engineers, the people whose job is keeping software running 24 hours a day. Those teams are drowning. AI is letting developers ship code faster than ever, which means more changes, more risk, and more pressure on the engineers watching the back end.
Empirik's pitch is that it handles the routine scanning and troubleshooting so those engineers can focus on harder problems. Chandrayana draws a direct comparison to Cursor and Claude Code, AI coding assistants that help software developers write code faster. "What agentic AI, meaning AI that can carry out multi-step tasks on its own, did for software," he told TechCrunch, "Empirik wants to do for infrastructure engineering."
For ordinary people, the upside is simpler: fewer website outages, fewer payment systems going dark, fewer hospital portals timing out at the worst moment.
What happens next?
Empirik already has paying customers across a wide range of company sizes, including financial data giant S&P Global and Guardant Health. The $21 million will fund growth as a fully independent company after two years inside Sequoia's walls.
Balkansky argues that no direct competitor occupies the same space yet, though Empirik sits alongside, rather than against, AI site-reliability platforms like Resolve and Traversal. The market for keeping systems alive is large and the cost of failure is high. Every minute a major retailer's checkout page is down costs real money, and right now most companies only find out something is broken once customers start complaining.
