Rockstar is recruiting for a well-funded seed-stage startup building the Agentic Execution System for global trade. Their agents don't assist with the work — they do it, reasoning over a live context graph of the business and executing actions with real financial consequence. They move real goods and real money for some of the largest enterprises in the world. The company is backed by leading VCs in NYC and SF, and the team brings deep experience across supply chain, financial technology, and applied AI.

Location: San Francisco — On-site
Employment Type: Full Time

About the Company

AI has learned to write code, draft contracts, and answer questions. It hasn't yet learned to run the physical world.

Global trade is the $10T proof of that gap. The physical world is a complex web of supply chains. Everything you own crossed oceans and borders to reach you, coordinated almost entirely by people. The data is too messy and too fragmented for traditional software to meaningfully automate, so the work stays manual and the cost of every mistake surfaces months too late.

The company is building the Agentic Execution System for global trade. Their agents don't assist with the work. They do it, reasoning over a live context graph of the business and executing actions with real financial consequence. They move real goods and real money for some of the largest enterprises in the world.

The company is backed by an oversubscribed seed round from leading VCs in NYC and SF, and the team brings deep experience across supply chain, financial technology, and applied AI.

The Role

As an early AI Engineer, the engineer owns the core of the system: the agentic harness that handles tool routing, policy, and memory, and the context graph it reasons over. The central challenge is reliability. These agents act on real money, on top of data that's messy, contradictory, and spread across many parties, and an agent that's right 95% of the time isn't good enough. Much of what this requires hasn't been solved anywhere yet.

The role comes with real product ownership. The engineer ships to production, works directly with how enterprise customers operate, and shapes what the platform becomes. The scope looks more like an early technical founder's than a spec to implement.

What the Engineer Will Build

• The agentic harness at the center of the platform: tool routing, policy enforcement, memory, and grounding for agents that execute real work end-to-end.
• A context graph of the physical economy: ingesting unstructured, contradictory data from emails, documents, and legacy systems and resolving it into a single source of truth.
• Evaluation and reliability infrastructure that makes agent reasoning safe enough for financial actions.
• The technical foundation of APIs, interfaces, and infrastructure that scales with customer usage and product complexity.

What the Company Requires

• 4+ years of experience in high-quality engineering environments
• Hands-on experience building agent applications: tool calling, MCP, context engineering
• Strong proficiency in Python, TypeScript, or similar; comfort with AWS/GCP
• Judgment about model tradeoffs: the ability to build the most reliable, cost-efficient system, not just the most impressive demo

What the Company Values

• Experience with LLM post-training and evaluation
• Proficiency with Neo4j or other graph databases (or the drive to learn fast)
• Startup experience: the ability to think like a PM and an engineer in the same conversation
• High ownership, direct communication, and a bias for action
• Interest in recruiting and mentoring future engineers