Deployment Guide#

This guide will walk you through deploying OSMO, a cloud-native platform designed for robotics developers to manage all aspects of AI and robotics development, from compute resources to data storage.

What is OSMO?#

OSMO is an open-source workflow orchestration platform that solves the complexity of scaling robotics and AI development across heterogeneous compute infrastructure.

The Challenge#

As AI and robotics workloads grow beyond single workstations, teams face critical scaling challenges:

Multi-robot simulations requiring coordinated execution across distributed systems
Distributed training across GPU clusters with complex dependencies
Hardware-in-the-loop testing on edge devices (NVIDIA Jetson, AGX)
Synthetic data generation at massive scale
Fragmented tooling that doesn’t integrate across environments
Infrastructure complexity that slows development cycles
Non-portable workflows that need rewriting for different hardware

The Solution#

OSMO provides a unified platform with three core capabilities:

Unified Control Plane

A single interface for workflow submission, monitoring, and management across all your compute resources. Define your workflow once in YAML and execute anywhere.

Bring Your Own Compute

Connect any Kubernetes cluster as an execution backend for your workflow:

Cloud: AWS (EKS), Azure (AKS), Google Cloud (GKE)
On-Premises: Data centers with DGX or OVX systems
Edge: NVIDIA Jetson devices for hardware-in-the-loop testing
Hybrid: Mix and match across environments in a single workflow

Bring Your Own Storage

Integrate with your existing storage infrastructure:

Any S3-compatible object storage (AWS S3, MinIO, etc.)
Azure Blob Storage

What You’ll Deploy#

An OSMO deployment consists of two main components:

1. OSMO Service (Control Plane)

The central service that provides the API and UI for workflow submission, monitoring, and management. This includes:

API server for workflow operations
Web UI for visual workflow management
Data storage configuration
Workflow scheduling and lifecycle management

2. OSMO Operators (Compute Plane)

An OSMO operator is an agent that you will deploy on your compute cluster to register it with the control plane similar to plug and play model.

Note

OSMO does not need network access to your compute cluster. Your compute clusters can run anywhere behind corporate firewalls, in restricted networks, or across geographically distributed locations. When you deploy our backend operator, it will initiate outbound connections to OSMO for registration and works like a plug and play model.