Empirik launches with $21M to help predict IT outages before they happen

Empirik is launching as an independent company after raising $21 million in seed funding from Sequoia, Canapi, and Alumni Ventures. The startup was incubated by Sequoia in 2023 and is focused on helping infrastructure teams predict outages before they happen.

The company’s premise is straightforward: instead of waiting for systems to fail, infrastructure teams can use AI to understand how changes may ripple across complex environments.

Empirik tracks system changes and infers their potential effects across an organization’s infrastructure. The goal is to help prevent and resolve incidents by giving teams a clearer view of dependencies that existing observability tools may miss.

TechCrunch describes the product as acting like an autonomous “traffic cop”: allowing low-risk changes, setting guardrails around larger changes, and flagging the most dangerous updates for human review.

The idea came from Avon Puri, Sequoia Capital’s chief digital and information officer, and Sudheer Dhurjati, another Sequoia IT leader, as large language models began showing more potential. Puri previously spent more than a decade running infrastructure at Rubrik and VMware.

Sequoia later recruited Kartik Chandrayana, formerly chief product officer at Quantum Metric and a Salesforce observability vice president, to serve as Empirik’s CEO earlier this year.

Empirik’s pitch is that infrastructure engineering needs its own AI acceleration layer as software development moves faster. The company says it wants to do for infrastructure engineers what tools like Cursor and Claude Code have done for software developers: automate certain tasks so teams can work faster.

The startup has already brought on customers ranging from startups to several Fortune 500 companies, including S&P Global, Guardant Health, and a major consumer packaged goods company. Sequoia’s Bogomil Balkansky described Empirik as complementary to AI SRE platforms like Resolve and Sequoia-backed Traversal.