Staff Engineer

Sauna•San Francisco, CA

11h•Onsite

About The Position

This is a full-time, in-person role based in San Francisco (Presidio). The company is seeking an experienced builder who will partner with the CTO on architectural decisions. The role involves making critical architectural tradeoffs, designing systems for concurrent agent sessions, and defining reliability and observability standards for a growing team. The Staff Engineer will also review code, set technical direction, and ship code daily, taking ownership of the product's evolution at scale.

Requirements

8+ years as an IC engineer.
Experience as Staff or Principal at a company with real scale.
Experience building and operating systems that serve millions of users.
Knowledge of the failure modes of distributed state, real-time streaming, and multi-tenant isolation from production experience.
Recent, hands-on AI/LLM work (built agent systems, model routing layers, or tool-use architectures).
Understanding of why tool execution in an agentic loop is fundamentally different from request-response.
Fluent in TypeScript and productive in it on day one.
Desire for real ownership and real risk.
US work authorization (O-1 or J-1 visa sponsorship considered for exceptional candidates).

Nice To Haves

Strong opinions on agent runtimes (Durable Objects vs. actors vs. microservices) and deployment strategies for long-running processes.
Experience with Cloudflare Workers at scale.
Contributions to or maintenance of significant open source projects.

Responsibilities

Make hard architectural tradeoffs and own the outcome (e.g., Durable Object SQLite vs. shared Postgres for session state, Cloudflare Workers CPU limits vs. longer-running workloads, Single-tenant sandboxes vs. multi-tenant pools).
Design the system that handles concurrent agent sessions across integrations, an agent responding to an iMessage, checking Calendar, drafting a Slack reply, and updating Linear, all within one session with consistent state.
Define reliability and observability standards for a ~20-person team: SLOs, error budgets, tracing strategy, incident response patterns.
Review every significant PR.
Set technical direction without blocking velocity.
Ship code daily.
Build new systems as needed.