About Mixmax

Mixmax is the #1 product helping revenue teams work proactively by eliminating busywork and creating seamless customer experiences, used by thousands of sales and customer teams inside Gmail and their CRM.

In 2026 we rebuilt how engineering works around AI agents. We are a small, senior team that ships with agents doing most of the typing, and engineers doing the thinking: deciding what to build, designing the system, directing agents, and owning what goes to customers.

How we work

• There are no fixed feature teams. Every engineer sits on one bench, and small pods of 1–3 people (often an engineer, a PM and a designer) form around a prioritized brief, ship it end to end, and dissolve.
• Agents are the default way to build. You orchestrate several agents in parallel, give them context, and verify what they produce.
• Every project follows one contract. A living spec with acceptance criteria, an append-only decision log, a release brief for sales and CS, and an honest retrospective.
• Verification gates, not handoffs. An agent verifier checks every PR against the spec, and another checks every release on staging before customers see it. There is no QA handoff.
• Designs are code. Designers ship AI-generated, production-ready prototypes; you build against them.
• Autonomy with accountability. You decide how to build, which tools to use and when to ship behind a flag. Compliance, production safety and customer-data rules always apply.
• We talk in weeks and scope down until it hurts. Outcome over output; we don't release and forget.

What you'll do

You will own product and platform projects end to end, from a one-line brief to a feature customers use, with AI agents as your main way of building.

• Shape the work. Turn a brief into a spec with clear user need, acceptance criteria, technical plan, cost, security and instrumentation, together with product and design.
• Design the system. Make the architectural calls, keep them simple and scalable, and log the decisions that matter.
• Orchestrate agents. Run multiple coding agents in parallel, feed them the right context (skills, rules, repo cards and our internal MCP servers), and keep cost and token spend in check.
• Verify everything. Review agent output with a critical eye, write the Playwright and unit tests, and pass the PR and release gates yourself.
• Ship and watch. Release behind feature flags, set up Datadog monitors and product analytics, use Cleric.ai to investigate production issues, and confirm the feature actually works for users.
• Take your turn on support. Join the support rotation, handle incidents, and turn recurring problems into new briefs.
• Improve the system we build with. Contribute skills, patterns and learnings to our shared context repo so the next pod moves faster.

In your first 90 days you'll: set up your AI-native environment, ship a scoped project with a pod, take a support rotation, and contribute at least one learning or skill back to the shared repo.

About you

You already build with agents every day, and you think like a systems designer and product owner more than a ticket-taker.

You'll likely have:

• 6+ years building and running production software, with deep backend experience and the range to work across the stack.
• A working AI-native setup you can show us: Claude Code, Codex, Cursor or similar; with your own skills, rules, MCP servers or agent workflows, and examples of running agents in parallel.
• Strong judgment on agent output. You catch subtle bugs, security issues and architectural drift, and you know when not to trust a green test run.
• Experience designing distributed systems on AWS: event-driven flows, queues, idempotency, failure recovery and cost.
• Ownership of quality without a QA team: test strategy, CI, observability, feature flags and incident response.
• Product sense. You can explain why a project matters to customers, cut scope to the essentials and measure the result.
• Clear written communication in English. Specs, decision logs and async updates are how a distributed team stays aligned.
• Comfort with ambiguity and change. The way we work is still evolving, and you help shape it.

Our stack

TypeScript and Node.js (Express), MongoDB, AWS (Fargate, Lambda, SQS, EventBridge, DynamoDB), Terraform, React in an Nx/pnpm monorepo, Vitest and Playwright, GitHub Actions, Datadog and Cleric.ai (AI SRE) for observability, Linear and internal MCP servers that give agents context on our services and tooling.

Nice to have:

• Building LLM-powered product features: agents, tool use, MCP servers, evals.
• Building internal agent tooling: PR review bots, sandboxed agents, context services.
• Gmail, Google Workspace or CRM (Salesforce, HubSpot) integrations.
• SOC 2 or security-conscious engineering in a SaaS company.

Why this role

• You work the way engineering is heading. Agents, verification gates and specs as code are how we ship today, not a pilot.
• Real ownership. You choose the work from a prioritized list, decide how to build it, and see it through to customers.
• Small team, direct impact. No layers between you and product, design, CS or leadership.
• Shape the model. Our operating model is young; your learnings and skills become how everyone works.

Location: remote, ideally LATAM (Brazil or Argentina preferably). If Europe, has to work Brazil/Argentina time zone. We can't hire in the US at present.