Software Engineer, Infrastructure

About Keycard At Keycard, we're building identity and access infrastructure for the agent-native era, where software isn't static but a dynamic, constantly changing system of AI agents working on behalf of people and businesses. Keycard gives every agent a verified identity and issues short-lived, task-scoped credentials at runtime, so agents can only do what they were asked to do, and every action is fully auditable. We're backed by a16z, Boldstart, and Acrew, and our founding team built and scaled Snyk, Auth0, HashiCorp, and Heroku.

The Role We're hiring two staff-level Software Engineers, Infrastructure to own how Keycard is built, shipped, and run.

Keycard is a control plane for agents: identity, credential exchange, delegation, policy, and audit. Many of the enterprises adopting it need that control plane inside their own cloud, behind their network controls, with no standing access for us. Bring-your-own-cloud (BYOC) is a product at Keycard, not a services engagement. It's versioned, documented, and held to the same reliability and security bar as our hosted service.

You'll own Keycard's build and release systems, our multi-tenant cloud, dedicated deployments, and deployments in customers' cloud accounts. Build common delivery and operating tools across them, with release safety, reliability, cloud cost, and repeatable installs as priorities. You'll work with founders, engineers, product, and customers, and agree escalation and on-call with engineering. We build on AWS (and later GCP and Azure), EKS, Pulumi, and Helm.

What You'll Do

BYOC product.

Own what it's like to install, upgrade, and get support for Keycard in a customer's cloud, from what their security team reviews to how each version is documented.

Compatibility.

Decide what we support, from clouds, regions, and Kubernetes versions to customer guardrails like SCPs and egress proxies. Set the skew limits and deprecation policy that keep our control plane and customer-hosted deployments compatible.

Deployments.

Make a new deployment, ours or a customer's, a configuration change: infrastructure as code, account bootstrap, and network patterns like PrivateLink and certificates that work through customer proxies. For example: take a new customer's AWS account to a running, upgradeable deployment with no standing access from us.

Releases.

Own one pipeline from commit to running deployment, wherever it runs, with the control plane kept separate from the data plane. Get releases to many independently versioned deployments with staged promotion, approval gates, and rollback.

Reliability.

Set the availability and durability targets we hold ourselves to for the hosted service, with the observability, runbooks, and incident response behind them. Operate customer-hosted deployments without standing access, using audited break-glass.

Agent operations.

Build agents for the repetitive work: pre-install checks, installs, failed-upgrade triage, and runbook drafts. They run with scoped credentials, every action is audited, and an engineer approves any change to a customer account.

Cost per install.

Keep each install cheap in engineering time and cloud spend. Decide what's standard, what's configurable, and what we decline, so one customer's request doesn't become the roadmap.

Deployment guardrails.

Work through reviews with customers' cloud and security teams. With security, set the guardrails for every deployment, from credential scope and isolation to workload identity and PKI, then build and run them.

What We're Looking For

You build infrastructure as software in Pulumi, Terraform, or CDK, and write the Go or TypeScript tooling around it. You've run Kubernetes and Helm in production.

Deep AWS experience, especially IAM and permission boundaries, Organizations and SCPs, and VPC networking.

You've run software in accounts you don't control, like customer VPCs, on-prem, or regulated networks, and supported it through upgrades.

You've shipped versioned releases that customers upgrade on their own schedule, with a support matrix and deprecations they could plan around.

You debug across layers: a hung request that turns out to be a security group rule, a proxy that breaks a TLS handshake, a certificate challenge that can't reach the internet.

You've worked directly with customers' platform and security teams, and turned down a one-off when a product change was better.

You think in control plane and data plane: you've designed or run systems where the two are separate and upgrade independently.

Agents at work.

You already use coding agents in infrastructure work and have a clear view of where they're reliable and where a human has to review.

Strong Candidates May Also Have

Experienc