Overview
You build and continuously improve AI agent systems that run Rocket's operations - marketing automation, growth workflows, internal tools, analytics pipelines, and research systems. The company already runs on AI agents. Your job isn't to introduce AI - it's to make existing systems work better and automate whatever's still manual. You work directly with the CEO and have full autonomy to identify problems and ship solutions.
Role Snapshot
| Aspect | Details |
|---|---|
| Role Type | AI/Automation Engineer (Rev Ops adjacent) |
| Primary Focus | Internal systems optimization and automation |
| Scope | Marketing ops, growth, internal tooling, analytics, research |
| Autonomy Level | Complete - no Jira, no task assignments, self-directed |
| Reporting | Directly to CEO |
| Team Size | First hire in this function - you define it |
Company Context
Stage: Early growth (likely Series A/B based on 106 employees and 400K users)
Size: 106 employees
Growth: Launched months ago, scaled to 400K+ creators across 180+ countries, 500K apps built
Market Position: Category creator in AI-powered app development - competing with traditional no-code tools and developer platforms
Product: AI platform that converts plain-English prompts into production-ready multi-page apps with data models and logic
What You'll Actually Do
Time Breakdown
Building/Shipping (50%) | Optimizing Existing (30%) | Research/Experimentation (20%)
Key Activities
- System optimization: Take existing AI agent workflows (marketing automation, growth systems) and rebuild them with newer/better models or frameworks when tools improve - which happens monthly
- Problem identification: Walk around the company (virtually), talk to teams, find manual processes or broken workflows, decide what's worth automating
- Agent development: Build new automation systems from scratch - could be a research pipeline one week, an analytics workflow the next, an internal tool after that
- Tool evaluation: Test new agent frameworks (swarms, memory systems), compare models (Claude, GPT-4, specialized models), decide what to use when
- Maintenance and monitoring: Make sure your systems keep working as APIs change, models update, and company needs evolve
What a Week Actually Looks Like
- Monday: Notice the marketing team manually pulling campaign data. Spend 4 hours building an agent that automates it. Ship it.
- Tuesday: Existing lead scoring system is using GPT-3.5. Rebuild it with Claude Sonnet because it's better at reasoning. Test against last month's data.
- Wednesday: CEO mentions a problem in Slack. You decide if it's worth solving. If yes, you build it. No approval process.
- Thursday: New agent framework drops. Spend the day experimenting to see if it's better than what you're using. Probably rebuild something.
- Friday: Three systems you built last month need tweaking because usage patterns changed. Fix them. Start thinking about next week's projects.
The Honest Reality
What's Hard
- No structure: There's no roadmap, no backlog, no one telling you what to work on. Some people freeze without structure. You need to wake up and decide what matters.
- Constant rebuilding: You'll build something that works great, then a better model drops two weeks later and you have to decide if it's worth rebuilding. The work is literally never done.
- Ambiguous impact: Sometimes you'll spend a week building something and realize it doesn't matter. No one's going to tell you if you're working on the right things.
- Tool overload: New AI tools and frameworks drop constantly. You have to filter signal from noise and not get distracted by every shiny new thing.
- First-hire isolation: There's no team, no peers in this function, no one to sanity-check your approach. You figure it out alone.
- Scope creep: Everything could be automated or improved. Saying no to good ideas is harder than saying yes.
What Success Looks Like
- Teams stop doing manual work you automated - they don't even remember the old way
- Systems you built months ago still run reliably without constant babysitting
- You ship 2-3 meaningful automations per month that actually get used
- When a new model or tool drops, you know within 48 hours if it's worth adopting
- The CEO stops thinking about operational problems you've already solved
Who You're Working With
Internal Stakeholders:
- CEO/Leadership - they throw problems at you, you decide what to solve
- Marketing team - they need campaign automation, lead enrichment, content workflows
- Growth team - they need analytics pipelines, experiment tracking, data systems
- Product team - they need internal tools, research automation, user insight systems
What They Care About:
- Speed: Can you ship something this week, not next quarter?
- Reliability: Does it keep working or break constantly?
- Practical impact: Does it actually save time or is it a cool demo?
Tech Stack (Inferred)
- Agent Frameworks: Likely using LangChain, AutoGPT, or custom agent systems
- Models: Claude (Code, Sonnet), GPT-4, possibly specialized models
- Tools: They mention Claude Code and Codex specifically
- Infrastructure: You'll be working with APIs, webhooks, automation platforms
Requirements
- 2-3 years building and shipping real AI/automation systems - not consulting, not prompting, actually building production systems
- Deep opinions on when to use which model, which agent framework, which tool - formed from actual experience
- Track record of side projects - they specifically want someone who can't stop tinkering
- Experience with agent swarms, memory systems, and modern AI development tools
- Comfort with zero structure - you need to self-direct completely
- Portfolio of use cases you've built (required for application - they don't want resumes)
Application Notes
- Don't send a resume - they explicitly don't want it
- Send 3 best use cases you've built via DM to Vishal Virani
- Video walkthroughs of each use case will jump you to the front of the line
- They want to see what you've actually shipped, not what you can talk about
Why This Role Exists
Rocket is an AI company that uses AI internally - they're eating their own dog food. They've already built agent systems across the company. Now they need someone to make those systems sharper every day and keep building new ones. This isn't an "AI strategy" role or a "let's explore AI" role. The infrastructure exists. You're making it better and expanding it.
The CEO is technical enough to build this stuff himself but doesn't have time. He wants someone who thinks like him - finds a problem, picks the right tool, ships it, moves on. If you need a PM to write specs or a meeting to get alignment, this isn't it.