Founding Engineer - Automation Platform

Most engineers build products. You’ll build companies. Welcome to Rocketable, where we’re creating fully autonomous software companies that operate at superhuman levels.

$/yr - $/yr

You’ll be architecting the nervous system of an entirely new kind of organism—a multi-agent AI platform that can run an entire portfolio of SaaS businesses without human intervention.

Software companies that operate at performance levels humans simply cannot achieve, enabled by intelligence that is distributed, scalable, and self‑improving.

• Week 1:
  You walk into your laboratory—a live SaaS business ready for radical transformation. It’s the perfect testing ground: real customers, real revenue, real complexity. Customer support takes hours. Bugs sit in backlogs. Features ship when someone remembers to deploy them. This is your canvas for proving what AI can really do.
• Week 4:
  You’ve deployed your first agent swarm. Customer support tickets are being triaged by AI. The meta‑agent layer is analyzing where humans still need to intervene, learning from those patterns. The orchestrator is discovering which agents work best together. The system is beginning to understand its own limitations.
• Week 12:
  The platform is running autonomously for hours at a time. Agents are spawning other agents to handle specialized tasks. The system has identified performance gaps from its data and built new tools to close them. Performance metrics are on a clear trajectory to exceed human baselines. You’re the conductor of an orchestra of AI systems working in concert.

• You’re a self‑directed learner who gets excited, not overwhelmed, by the pace of AI advancement.
• You’re comfortable throwing away three weeks of work because a better approach just became possible.
• You thrive in high ambiguity—there’s no Stack Overflow answer for what we’re building.
• You see “no playbook exists” as an opportunity.
• You believe full automation for software companies is inevitable.

• Orchestrator‑worker patterns: dozens of specialized agents, parallel coordination, distributed queuing.
• Cross‑stack abstractions: works across any portfolio company’s tech stack and schemas.
• Meta‑agents learn from overrides: analyze why humans intervene, accelerate improvement.
• Replay thousands of scenarios: test new models against history, catch regressions.
• Agents create own tools: programmatically generate solutions when stuck.
• Time‑travel debugging: structured logs, audit every decision, full observability.
• Human approval gates: security boundaries for code execution, data mods, critical ops.
• Hybrid data fusion: databases, APIs, event streams – traditional + vector stores.

• You’ve built and scaled production systems handling 100K+ daily active users.
• Deep experience with distributed systems: microservices, event‑driven architectures, message queues.
• Full‑stack development: you can debug everything from React components to database queries and underlying infrastructure.
• Fluent in multiple programming languages (Type Script and Python preferred).
• Track record designing APIs that other developers actually want to use.

• Experience integrating LLMs via foundation model APIs (OpenAI, Anthropic, Google).
• You understand prompt engineering as an engineering discipline, not just trial and error.
• You’ve built eval systems to measure AI performance.
• Experience with token optimization, context window management, and fallback strategies.
• Bonus:
  You’ve worked on self‑improving systems, reinforcement learning, or RLHF/RLAIF.

• Kubernetes expertise for container orchestration.
• Docker mastery: you understand security, resource limits, and ephemeral environments.
• Infrastructure as Code:
  Terraform or similar—your infrastructure is version‑controlled and reproducible.
• Cloud platform expertise (GCP preferred, AWS/Azure also OK).
• You’ve built CI/CD pipelines that actually accelerate development instead of slowing it down.
• Production monitoring and observability, including debugging distributed systems.
• Security fundamentals: sandboxing untrusted…