Skip to content
ResortOS

A proprietary ecosystem byResortOS

RAVENAn intelligent technology ecosystem

Engineered by ResortOS, Raven brings together quantitative research, intelligent analysis, coordinated software systems and advanced automation within a unified technology architecture.

Data IntelligenceQuantitative ResearchAnalytical IntelligenceAutomation SystemsRisk and MonitoringOrchestration and Control

What Raven is, and why it exists.

Raven is a proprietary software ecosystem that ResortOS designs, builds and runs for its own use. It brings data, research, analysis, automation, risk control and monitoring together, so that decisions are made from shared data, tested ideas and written limits rather than instinct.

It is built as many independent components rather than one large program. Each has a narrow job and a clear contract with the rest: a data hub that every component reads from, specialised engines with their own workflows, a coordinator between them, a risk layer that checks every action, and interfaces that keep people in charge.

That structure is the point. Components can be tested, changed or switched off one at a time, ideas are proven before they are used, and the whole system stays visible while it runs. Raven is developed continuously, one tested change at a time.

Six layers, one architecture.

Data flows up from a shared hub into research and analysis; tested changes and context reach the engines; every action passes the risk layer; and the control layer watches it all. The layers are how we describe the system; each one maps to components that exist.

Layers are shown top to bottom, from control down to data. Select one to see how it connects.

Layer

Automation Systems

Independent engines run their own workflows, while a coordinator makes sure they never work against each other.

  • Several independent, specialised engines
  • A coordinator that allocates work between them
  • Scheduled and event-driven workflows
  • Execution infrastructure with recovery after restarts
  • Release tooling with checks before every deploy

Receives

  • Only tested changes are released
  • Context for each decision
  • Every action checked against limits

Feeds

  • The people operating the system

Components in this layer

The components of the ecosystem.

Names used to describe Raven publicly. Each maps to working software; they are parts of one internal system, not separately sold products.

  • Raven Core

    System architecture and coordination

    The foundation every other part runs on: a shared message bus, durable storage, the data hub and a crash-safe cache that keep components in step.

    Every component talks through Core.

    Data Intelligence

  • Raven Intelligence

    Analytical and intelligent processing

    Analysis of changing conditions, contextual evaluation, and a locally hosted language model used for advisory work under strict, checked answer formats.

    Feeds context to Automation and evidence to Research.

    Analytical Intelligence

  • Raven Research

    Quantitative research and evaluation

    A separate research workspace for modelling, historical replays and pre-registered evaluations, with shadow books that run ideas forward safely.

    Decides what Automation is allowed to change.

    Quantitative Research

  • Raven Automation

    Automated workflows and coordinated engines

    Several specialised engines, each with its own workflow, run alongside a coordinator that allocates work so they never conflict.

    Acts only within the limits Risk enforces.

    Automation Systems

  • Raven Risk

    Risk analysis and monitoring

    Written risk limits checked before each action, circuit breakers, activity monitoring, reconciliation and scheduled audits.

    Sits between every engine and the outside world.

    Risk and Monitoring

  • Raven Control

    Monitoring and operational interfaces

    Raven's own control centre: a web dashboard and a mobile app, with alerts and approvals for actions that need a person.

    Gives people visibility of every layer.

    Orchestration and Control

  • Raven Staff

    Supervised companion application

    A companion application for supervised operation, where sensitive actions wait for a person's approval and every step is recorded.

    Works under the same Risk limits as the engines.

    Orchestration and Control

  • Raven Academy

    Model training and evaluation

    The environment where the locally hosted model is fine-tuned and examined, with every candidate scored before it is considered.

    Supplies Intelligence with tested models only.

    Analytical Intelligence

Engineering principles.

Built with

  • Python with asyncio
  • aiohttp and FastAPI services
  • PostgreSQL
  • Pydantic data contracts
  • Docker
  • React, TypeScript and Vite for the dashboard
  • A native iOS app
  • A locally hosted open-weight language model
  • Modular by design

    Independent components with clear contracts, so one part can be changed, tested or switched off without touching the rest.

  • Evidence before change

    Ideas are written down as tests before they run, checked on history, then run forward in shadow before anything live changes.

  • Data shared, not duplicated

    One hub collects and normalises data once, so every component works from the same picture.

  • Risk as architecture

    Limits are code, checked before every action, not a report someone reads afterwards.

  • Models held to account

    Model outputs are measured against results, kept to strict formats, and switched off where they do not add value.

  • Visible and recoverable

    Health checks, audits and a full record of events, with engines that pick up correctly after a restart.

How the system works together.

Controlvisibility, alerts, approvalsEngineown workflowEngineown workflowEngineown workflowCoordinatorallocates the workRisk gatelimits, breakersExternalsystemsData hubone shared picture
CoordinationIndependent engines take their work through a coordinator; a risk gate checks every action before anything reaches an external system; the control layer watches and holds approvals. Illustration of the structure, not a live interface.
Ideaa question worth testingWritten testpass and fail set firstHistory replaysimulated on past dataShadow runforward, nothing liveDecisionadopt, change or dropthe next question
Research loopHow an idea earns its way into the system: written as a test first, replayed on history, run forward in shadow, then adopted, changed or dropped. Illustration of the process.

What this page does not show

Raven is internal software. This page describes its architecture, not its operations.

  • No live data, positions, balances or results
  • No strategy rules, model prompts or configuration
  • No account, server or infrastructure details
  • No screenshots of private interfaces: diagrams on this page are illustrations

Built by ResortOS

One engineering team, more than one product.

Raven is part of ResortOS's broader engineering vision: developing proprietary software systems that combine practical engineering, intelligent data processing, quantitative research and advanced automation. The same team builds the ResortOS platform for resorts and hotels, and custom software for other businesses.

Raven is used inside ResortOS. It isn't offered as a public service.