Staff Product Engineer (São Paulo)

We're restructuring engineering around initiative teams: a Product Engineer paired with a PM and a designer, with an Engineering Manager who covers a couple of initiatives and supports your growth. The engineer leads AI agents like a team, ships the work, and is accountable — with the rest of the triangle — for whether the initiative moves its metric.

We're betting that 1–2 strong engineers running AI agents can outship the labor-team model that defined the last decade of software. That bet only works if the engineers we hire are wired for ownership and can ship to a marketplace with real customers and pros on both sides.

The Role

You're the engineering anchor of one initiative at a time. The initiative is a team effort — an iron triangle of you, your PM, and your designer — and you have key participation across the full lifecycle: shaping the problem, deciding the technical approach, leading the AI agents that implement most of the code, shipping to production, and answering for the outcome alongside the rest of the triangle.

You're accountable for the outcome — not for the volume of code merged. If an agent can ship it safely, your job is to make sure the agent does it right and the metric moves. If the initiative needs hand-written code in a sensitive area, you write it yourself.

What makes this role different

• You lead AI agents, not humans. Claude Code, Cursor, Codex, and our internal agent stack are your team. You own the quality, safety, and velocity of what they produce.

• You own an outcome, not a ticket queue. Problem-framing through production through the metric review 2–4 weeks after launch.

• You partner horizontally with PM and design. No tech lead above you. No architect approval. No ticket grooming committee.

• The bar is staff, not senior. You make the call when the call needs to be made. If you're waiting to be told, this isn't the role.

What You'll Own

• The technical approach — architecture, data model, integration choices, rollout plan, observability, and rollback strategy for your initiative. You make the call, document it, and revisit it if the data says you were wrong.

• Agent-led implementation quality — the prompts, guardrails, evals, tests, and review loop that let agents ship safe, correct, production-ready code on your initiative. Most lines will be agent-authored. You're accountable for them.

• Cross-functional partnership — daily working contact with your PM (scope, tradeoffs) and your designer (UX decisions, in-tool prototyping with agents), and weekly check-ins with your EM (initiative health, blockers, growth).

• The initiative outcome — the specific metric the initiative was set up to move. In partnership with your PM, you present results 2–4 weeks post-launch and share the "did it work" answer.

• A high bar for what ships under your name — production correctness, security posture, performance, observability, and the experience for customers and pros. Agents accelerate you; they don't lower the bar.

Problems to Solve

Leading AI agents at staff-level quality Most of the code on your initiative will be authored by AI agents. The work is making agents ship as if a senior engineer wrote it: prompts that encode our codebase conventions, evals that catch hallucinations before merge, tests that exercise the edges, observability that catches the regression in production before a customer reports it. How do you build the agent workflow that lets one engineer ship what used to take a team?

Owning an outcome without a tech lead You don't have a tech lead to approve your design or an architect to escalate to. You have an EM who covers a couple of initiatives and peers on adjacent ones. How do you make calls fast, document them clearly, and stay accountable to the outcome — without slowing down for hierarchy that no longer exists?

Shipping outcomes, not features The initiative will be measured by a metric — a conv