Quantitative systems · AI-assisted engineering
Nebula
From market data to an operable trading system.
Nebula brings the engineering around quantitative trading into a connected system. Its scope reaches beyond a strategy script: market data has to arrive consistently, research needs to be repeatable, execution must reflect operational state, and an operator needs a clear view of what the system is doing. The project connects these responsibilities through a reactive backend, research tooling and a browser-based operations interface.
- Java
- Spring WebFlux
- MongoDB
- Python
- GitHub Actions
Build the system around the research.
A research result and an operational application answer different questions. Research needs a consistent historical view and a way to repeat an experiment. An execution service needs to respond to changing external state, recover from interrupted work and expose enough information for an operator to understand what happened. Nebula addresses both sides within one engineering project.
The implementation includes streaming and historical data ingestion, backtesting and simulated execution, exchange execution services, reconciliation, health checks and dashboard views. These capabilities provide a foundation for investigating and operating the system while keeping the trading method separate from its surrounding infrastructure.
The engineering around the trading system
Data feeds research and execution capabilities. Monitoring and reconciliation provide an operational view; proprietary trading logic is omitted.
Data sources
- Historical dataRetained inputs for repeatable researchNext → Research & simulation
- Streaming dataIncoming market eventsNext → Execution services
System capabilities
- Research & simulationBacktesting and simulated executionNext → Operator dashboard
- Execution servicesExchange-facing executionNext → Recorded state
- Recorded stateApplication state and reconciliationNext → Health & monitoring
Operational visibility
- Health & monitoringInspect service and dependency stateNext → Operator dashboard
- Operator dashboardA browser view of the system
AI helps build the pipelines. Engineering checks govern the changes.
AI-assisted development is organized around planning, source review, implementation and audit. Reusable project context gives coding agents the constraints of the codebase, while task-level instructions call for verification and regression checks. This makes AI part of the engineering workflow for developing and refining the data, research and operational pipelines; responsibility for the system remains with its human owner.
The project also contains separate tooling for researching AI-assisted decisions. That work evaluates model outputs against retained evidence within a controlled research process. It is distinct from the use of coding agents to build software, and it is not presented here as evidence that an AI model autonomously controls production trading.
Where AI enters the engineering workflow
AI supports development of the pipelines. Review and verification remain engineering responsibilities; this is separate from live trading decisions.
Context & direction
- Scope the changeGoals, codebase context and constraintsNext → Plan & implement
AI-assisted development
- Plan & implementSource review and code changesNext → Checks pass?
- Checks pass?Tests and engineering reviewPass → Release workflowRevise → Plan & implement
Delivery
- Release workflowArtifacts, integrity and deployment checks
Make changes inspectable across the lifecycle.
Research and operational behavior can diverge as a system evolves. Nebula includes consistency checks and separates routine tests from more extensive research checks, allowing the two forms of validation to serve different purposes. Operational code includes reconciliation and health reporting so the application can compare its recorded state with the external systems it depends on.
Release and infrastructure workflows cover packaged artifacts, integrity checks, deployment health and recovery tooling. The dashboard brings the operational view to a browser. Together, these elements demonstrate the breadth of engineering required around a quantitative application: data work, backend development, verification, release processes and usable operations.
A complete platform capability, with the trading edge kept private.
Nebula demonstrates implementation across the lifecycle of a quantitative software system, from acquiring data and running research to execution infrastructure and operator visibility. The transferable capability is building connected software whose behavior can be inspected and whose changes can be evaluated systematically.
This overview intentionally keeps trading rules, research recipes, model instructions, configurations and financial results private. It describes engineering capabilities rather than making a claim about investment performance.
Building a data-intensive product?
Scope the data flow, backend capability or operational interface your team needs to deliver next.
Discuss your project