Empirik Raises $21M for Autonomous Infrastructure Engineering

Empirik emerged from stealth with $21M in Seed funding to build an autonomous infrastructure engineer, an AI system designed to understand proposed infrastructure changes before they reach production. The company says the financing came from Sequoia Capital, S32, Canapi Ventures, and Alumni Ventures.

The announcement puts capital behind a widening enterprise mismatch. AI coding agents can generate and ship software quickly, while the infrastructure underneath that software is still governed through human review, ticket queues, runbooks, and knowledge scattered across teams. Empirik is trying to give both engineers and agents a live model of the environment so they can calculate what a change will affect before it happens.

Empirik was incubated inside Sequoia Capital in 2023 and is now operating as an independent company. Its founders, Avon Puri and Sudheer Dhurjati, began with experience running enterprise technology at VMware, Rubrik, and Sequoia. Kartik Chandrayana joined as CEO after leading more than 30 observability, availability, and big-data products at Salesforce and later serving as chief product officer at Quantum Metric.

What Empirik raised

Empirik announced more than $21M in funding on September 1, 2026 through its company blog. TechCrunch described the financing as a $21M Seed round, and Kartik Chandrayana used the same Seed label in the launch announcement shared through Sequoia's LinkedIn activity. No valuation was disclosed.

The investor group includes Sequoia Capital, S32, Canapi Ventures, and Alumni Ventures. Public materials do not break down how much each investor contributed, identify a separately priced prior round, or provide a detailed allocation of proceeds. The safest capital accounting is therefore one $21M Seed financing, not a claim about a lead investor's check size or a larger lifetime total.

The official press release appeared on September 2, one day after the company blog and TechCrunch report. DevCuration uses September 1 as the event date because that is when Empirik publicly announced the financing and emerged from stealth.

Why infrastructure change is the product

Traditional observability starts with symptoms: latency increases, errors climb, a service becomes unavailable, or a pod crashes. Engineers then work backward through logs, metrics, traces, configuration histories, and human memory to reconstruct what changed.

Empirik wants to move that reasoning earlier. When an engineer or AI agent initiates an action from a pull request, ticket, or pipeline, the platform captures the intent and projects the proposed mutation across a continuously updated model of the environment. The model covers cloud infrastructure, on-prem systems, Kubernetes, virtual machines, identity and access management, CI/CD, and SaaS services.

The company's term for that model is an infrastructure compiler. It converts metadata from many systems into an operational graph that maps assets, accounts, dependencies, and changes. Empirik says the graph can help teams assess blast radius, identify affected owners, detect drift between intended and live state, and block or flag risky actions before deployment.

This is more than a new dashboard. The commercial proposition is that a sufficiently complete operational map can become a control layer. A low-risk change might proceed automatically, a consequential change might require governed approval, and a dangerous change might be stopped before it creates an incident. That progression depends on whether the map is accurate enough for customers to trust the decision.

The people behind Empirik

Avon Puri and Sudheer Dhurjati developed the initial version inside Sequoia after seeing the same infrastructure problem across several large technology environments. Empirik's current company page lists Puri as co-founder and board director and Dhurjati as co-founder and CTO.

Puri previously held senior technology roles at VMware and Rubrik before becoming Sequoia's first global chief digital officer. Dhurjati previously served as vice president of technology at Sequoia and has worked across cloud infrastructure, Kubernetes, distributed systems, and observability.

Chandrayana now leads the company as CEO. He previously co-founded Twin Prime, which Salesforce acquired in 2016, then spent five years leading Salesforce products across observability and big data. He later served as chief product officer at Quantum Metric. The leadership mix gives Empirik founders who experienced the infrastructure problem internally and a CEO who has built and sold software into enterprise engineering teams.

Early production evidence

Empirik says it is already in production at Guardant Health, Avahi Systems, TCBPay, an unnamed Fortune 50 consumer-products company, and an unnamed Fortune 500 financial-data-services company. TechCrunch independently reported Guardant Health and a large consumer-products customer, while the named performance details come from the company and Sequoia.

Sequoia reports that TCBPay, a payment processor handling more than $1.5B in annual volume, reduced detection time for sensitive configuration changes from 30 minutes to 10 seconds. It also says Guardant Health expanded Empirik from production into development and on-prem environments, and that the platform processes millions of weekly change and telemetry events across dependency graphs covering a few million resources.

Those figures are meaningful evidence of deployment, but they are not an independently audited measurement of reliability, revenue, retention, or incident reduction. They show where the product is being tested and expanded. The longer commercial record will depend on repeatable outcomes across heterogeneous customer environments.

What the $21M changes

The financing gives Empirik more capacity to expand its product and enterprise adoption as AI-generated code increases the frequency of infrastructure change. It also raises the standard the company must meet. A system that only understands part of an environment may miss the dependency that turns a routine deployment into a production event.

Enterprise buyers will therefore judge Empirik on coverage, accuracy, governance, integration depth, and the confidence required to let software act. The product has to connect declared configuration with the state that is actually running, preserve context across legacy and modern systems, and explain its reasoning to the people accountable for uptime and security.

The broader market signal is that agentic software is creating demand for new infrastructure controls. Coding assistants made generation faster. The next enterprise layer will help decide which machine-generated changes can proceed, which need a human, and which should never reach production. Empirik's $21M Seed is a wager that operational context becomes the permission system for that transition.