We are a technology company that happens to trade. Our engineers don't support the business—they are the business.
class ExecutionEngine {
public:
void optimizeLatency(NetworkPacket& pkt) {
// Daily ingest is 10GB. The historical store is measured in petabytes.
// Execution latency is under 1 millisecond.
}
}
Daily ingest is 10GB. The historical store we replay is measured in petabytes. Execution latency is under 1 millisecond.
The systems are ours. A signal is replayed before it is trusted.
We are agnostic to tools but religious about performance. We use the right tool for the job, whether it's bare-metal C++ or cloud-native Kubernetes.
Execution latency is under 1 millisecond. Orders go out through the prime broker and are sliced to limit impact.
The simulation engine replays history before a signal is trusted.
Daily ingest is 10GB, from the markets the book covers. The historical store we replay is measured in petabytes.
We operate with a "Zero Trust" architecture. Automated risk checks sit on the wire, ensuring that no bug can threaten the firm's capital. Observability is built into every service.
Engineers build the path from a research notebook to a tested signal: feature stores, evaluation, and deployment. Production changes go through peer review and the Gauntlet. Low-latency trading code and model tooling are different jobs, and both sit on this team.
"We came from aerospace, where iteration cycles took years. Here, we design a circuit in the morning, verify it by lunch, and see it trading in Tokyo by the evening. The feedback loop is addictive."
"Rozario isn't about maintaining legacy code. We are constantly rewriting our core stacks to leverage new hardware. If you want to work on the bleeding edge of C++, this is it."
"The historical store is measured in petabytes. A signal is replayed before it is trusted."
Research code stays with the firm. A signal is replayed before it is trusted.
# research_code.py
# Research code stays with the firm.
# A signal is replayed before it is trusted.
You never graduate from Rozario. We invest heavily in continuous education to ensure our team remains at the frontier of computer science.
Weekly seminars led by principal engineers on topics like 'Lock-free Queues' or 'Kernel Networking.'
Unlimited budget to attend KubeCon, CppCon, PyData, or any conference relevant to your craft.
Quarterly 24-hour events where engineers are encouraged to build wild, experimental tools.
We are looking for engineers who are obsessed with performance, reliability, and scale. If that sounds like you, let's talk.