About RoZetta

We build AI-native data infrastructure and intelligence products that help financial institutions, global exchanges and data vendors modernise their data, streamline operations and launch new data products faster.

The Role

At RoZetta, AI agents write most of our code. We are looking for an engineer to design the system those agents work in: the skills, the architecture, the tests and the evals that turn agent output into production software for global financial markets.

What you will do:

- Deliver features end to end by orchestrating teams of AI coding agents: orchestrators, parallel workers, reviewers and evaluators.
- Design and maintain the agent skills, context, tools and MCP servers that every engineer and agent on the team uses.
- Characterise systems before you change them: boundaries, data flows, invariants and failure modes.
- Set the architecture principles and design patterns our agent teams follow: modular design, clear contracts, domain-driven design.
- Build the tests and evals that let agents move fast safely, and make CI the gatekeeper.
- Build AI features into our products, with evals, tracing and cost measured from day one.
- Work with our Forward Deployed Engineers to turn client solutions into reusable platform capability.
- Help the team make the shift by pairing, reviewing and sharing what works.





You are a good fit if you have:

- Senior full-stack engineering background with real data depth: Python, TypeScript, SQL, data modelling and AWS. Ideally, some exposure to building production data pipelines or to data governance (quality, lineage, access control).
- Shipped production software substantially built by AI agents, and can show us the workflow, the guardrails and where it breaks.
- Robust systems thinking: you can map an unfamiliar system (its boundaries, data flows and failure modes) and explain what you would change and why. You have owned system architecture: modular systems, contracts and trade-offs.
- A real testing discipline. Your tests have caught what agents got wrong.
- A habit of starting things nobody asked for, and finishing them.

Nice to have

- Designed agent skills, prompts or tooling that other engineers use
- LLM evals and tracing for AI features using Langfuse
- Security for agents working near sensitive data
- Iceberg, ClickHouse or Spark; React; Kubernetes
- Financial markets or financial data

Why Rozetta

- Work agent-first, on a platform designed for it.
- Real ownership. Your patterns become how the whole team builds.
- Hard, interesting problems in global financial data.

📌 Agentic AI Engineer (Sydney)
🏢 RoZetta Technology
📍 Sydney