Auraison Platform
What is Auraison?
Auraison is an agentic platform for mission-critical applications such as robotics, perception and others. It is centered around the orchestration capabilities of coding agents and allows users to transform general agents into domain-specific agents that offer a 10x advantage both in terms of performance as well as operational costs. It uses cutting-edge technologies such as agentic reinforcement learning, hierarchical planning and yield management while at the same time its open architecture can interface seamlessly to any model or data repository and any environment/runtime. Applications can be deployed to your Cloudflare account.
Auraison's functionality can be divided into four-planes:
| Plane | Responsibility | Key Components |
|---|---|---|
| Control plane | Access, Jobs, Agent Lifecycle, Experiment Tracking | CLI, FastAPI API, Claude Managed Agents, Pi Agents |
| User plane | Execution on Ray clusters and Cloudflare Workers | torch.dev.gpu, ros.dev.gpu Ray environments |
| Data plane | Persistent storage, context retrieval | Lakehouse with DuckDB query engine and DuckLake |
| Management plane | Deployment, Billing, Tenancy, Quotas, Cost Control |
See architecture/system-architecture.mdx for the full design.
Reference Applications
The platform provides the substrate on which reference applications run. The platform itself is application-agnostic but we demonstrate here various domain-specific applications to illustrate its flexibility.
| Application | Domain | Repository | User Plane | Status |
|---|---|---|---|---|
| turtlebot-maze | Mobile robot navigation | aegean-ai/turtlebot-maze | ros.dev.gpu | Active |
| ar4-physical-ai | Robotic arm manipulation (VLA) | aegean-ai/ar4-physical-ai | ros.dev.gpu + torch.dev.gpu | Active |
| Deep Evidence Agent (DEA) | Engineering traceability & evidence-grounded reasoning | aegean-ai/dea | torch.dev.gpu | Design phase |
| counter-uas | Aerial object detection & tracking (COCO / VisDrone / BDD) | — | torch.dev.gpu | Planned — datasets in lakehouse |
| usv-vrx | Maritime USV navigation (Virtual RobotX) | pantelis/usv-vrx-2022 | ros.dev.gpu | External repo — integration pending |
| sidewalk-extraction | Geospatial sidewalk segmentation from aerial imagery | — | torch.dev.gpu | Planned — datasets in lakehouse |
| anomaly-detection | Industrial surface-defect detection (DAGM) | — | torch.dev.gpu | Planned — dataset in lakehouse |