Infrastructure Engineer

Build and operate the secure execution substrate for enterprise agents, customer applications, and Sycamore's control plane.

Sycamore is building the trusted agent operating system for the enterprise. Our platform helps companies build, deploy, and orchestrate agentic apps that take on real operational work, with the security and control large organizations need.

We are a small, engineering-led team working directly with Fortune 500 enterprises. We have raised $65M from Coatue and Lightspeed, along with other investors and industry leaders.

This is software engineering for consequential distributed systems, not cloud administration: control-plane services and Kubernetes controllers, identity and network boundaries, tenant-isolated data, and failures that cross application, cluster, database, and customer-network layers. Infrastructure is the foundation beneath Product and Core AI, so an error here can affect every application and customer. The team covers three areas. They share an on-call rotation and a great many failures, so nobody works in only one.

The platform and control plane. Kubernetes-based serving, the sandboxes agents execute in, deployment and rollback, networking and identity, and the operators that reconcile all of it. This is where blast radius is decided.

The data platform. This is database platform engineering, not analytics: there is no warehouse, no dbt, and no pipeline waiting for an owner. What we have is a growing fleet of PostgreSQL databases carrying enterprise customer data under a genuine isolation requirement, hundreds of migrations across many trees and authors, a connection budget that is a real ceiling, retention obligations measured in years, and a hybrid full-text and vector retrieval path in the hot path of every agent conversation.

The problems are fleet-shaped: a migration is easy, and applying it safely across every tenant database with per-tenant failure isolation and no downtime is not.

Reliability. How the platform behaves over time: service level objectives and error budgets that mean something, an observability estate managed as code, promotion gates that decide whether a release reaches production, capacity and cold-start performance, and incident response from page through postmortem to the structural fix. We have a large monitor fleet and an enforced metric taxonomy, but few service level objectives, no error budgets, no burn-rate alerting, and a promotion gate that warns rather than blocks.

Building that practice is the work, not a side project within it.

• Build and operate Kubernetes-based serving and sandbox platforms for agents, internal services, and customer applications.
• Develop infrastructure-as-code, Helm charts, provisioning services, admission controls, and operators that reconcile environments safely, along with the change controls, drift detection, release gates, and fail-safe rollout procedures around them.
• Own the database fleet: schema topology and migration safety, provisioning and credential rotation, pooling and connection economics, isolation guarantees, retention and partitioning, and the search and embedding path.
• Define service level indicators and objectives, establish error budgets, and own the observability estate as code: monitors, dashboards, tracing, metric taxonomy, burn-rate alerting, and the policy that keeps it from decaying.
• Lead incident response from detection through containment, root cause, postmortem, and the structural fix.
• Support hosted and customer-controlled deployment patterns while keeping the operational model consistent and auditable.
• Travel to customer sites when working alongside a customer's engineering or security team will materially improve a deployment or incident outcome.

Our current infrastructure environment includes GCP, Kubernetes and GKE, Knative, Envoy, Cloudflare, Pulumi, Helm, Build Kit, PostgreSQL and Cloud SQL, object storage, container registries, workload identity, Python and Go control-plane services, Kubernetes operators, Git Hub Actions, and production observability.

Alert triage is substantially agent-driven, with automated investigation, ticket filing, and a decay process that nominates monitors nobody has acted on for deletion. We also support customer-controlled deployment requirements and maintain more than one application-serving path during platform migrations.

This is context, not a…