Partnering with empirik: Building the autonomous infrastructure engineer
Change has always been the fundamental operational signal in infrastructure. empirik is turning that signal into the foundation for autonomous, agentic infrastructure engineering.
I spent years at Salesforce running more than 30 infrastructure products, living the same tradeoff every day: move fast and take on risk, or slow down to protect reliability and stall the business.
For years, infrastructure teams managed that tension with ticket queues, change review boards, runbooks, and observability after the fact. It was imperfect, but it worked because humans were still producing change at human speed.
That assumption is now breaking.
Software is becoming autonomous. Infrastructure is still manual
Coding agents are already writing, refactoring, and shipping software at machine speed. The result isn't just more code—it's an explosion in infrastructure change: more deployments, shifting permissions, configuration updates, provisioning, and fluid dependencies.
Yet the infrastructure operating model hasn't changed. Humans still manually review changes, guess dependencies, estimate blast radius, and respond after systems fail.
“You cannot put fast-moving agentic software on manually operated infrastructure and expect the system to hold.”
Adding headcount won't fix this. Modern observability diagnoses issues after the fact, and AI SRE merely accelerates incident response. Neither prevents failure. To prevent incidents, systems must understand what a change touches before execution.
Software has a repo. Infrastructure does not
Coding agents succeeded because the codebase is a structured, versioned, and machine-readable artifact they can reason over before acting. Infrastructure lacks this foundation:
- IaC defines declared state, not actual resource relationships. It drifts constantly, and critical dependencies never appear in code.
- Service meshes and runtime topology only see what communicates, missing virtual constructs such as IAM roles, routing tables, quotas, and security groups, which are key to understanding infrastructure.
- CMDBs require constant mapping and maintenance, regularly fall behind, and miss ephemeral cloud and Kubernetes workloads altogether.
- Point tools understand their own domain, but not how cloud, SaaS, code, CI/CD, IAM, Kubernetes, and on-prem systems connect.
Without that shared model, autonomous infrastructure is guesswork. And in infrastructure, a guess has a blast radius.
empirik builds the missing operational layer
Every infrastructure action has two sides. There is intent: what someone or some agent is trying to do. And there is impact: what that action will actually touch.
empirik captures engineer's intent, computes blast radius, and safely evaluates whether a change can execute autonomously or needs human review.
Change Events as Primary Signal
Treating change events as the primary signal before execution, rather than diagnosing failures after runtime.
Unified Infrastructure Compiler
Constructs a live, true model across on-prem, cloud, and SaaS; elevating virtual constructs like IAM policies and routing tables to first-class citizens.
Understand intent. Compute consequences. Then safely act.
From reactive operations to proactive infrastructure
Enterprises are using the same operational model today to move infrastructure work upstream:
- Block or flag risky changes before they merge or deploy.
- Assess blast radius and identify affected owners before a change lands.
- Detect drift between intended and live state.
- Enrich change governance with the actual infrastructure affected.
- Give engineers and AI agents the same ground-truth context for architecture, migrations, incidents, security, and provisioning.
The same system that answers questions today can validate actions next, and safely execute them as enterprises grant it more authority.
Infrastructure's next chapter: Autonomous Infra Engineer
Software has crossed the threshold into machine-speed creation. Infrastructure can no longer be governed by human-speed reasoning.
empirik is building the autonomous infrastructure engineering layer to bridge this divide.
Today, we are coming out of stealth with over $21M in funding led by Sequoia & S32 alongside Canapi Ventures and Alumni Ventures. More importantly, empirik is already in production—powering enterprise environments across Guardant Health, a Fortune 50 CPG leader, a Fortune 500 financial services leader, and more.
Bring us one environment.
We will build the operational model and show you a dependency you did not know existed.
Run into something we haven't named?
empirik is provider-agnostic. If a system declares infrastructure or records change, it can become part of the compiled graph.