Infrastructure & developer tools
Built for developers. Ready for enterprise.
Use Oumi from the CLI, Python API, and AI coding tools, with the security and infrastructure controls required for production AI.
Built on Oumi’s open-source foundation, the Oumi platform combines developer flexibility with enterprise access controls, compliance, and private deployment options.
PRIVATE INFRASTRUCTURE
Run Oumi inside the infrastructure you control.
Keep your data and AI workloads inside your own environment while using Oumi to build, evaluate, train, and operate models and agents.
Private VPC
Run Oumi workloads inside your cloud environment with control over networking, data, compute, and access.
On-premises
Run Oumi workloads on your own Kubernetes and GPU infrastructure when data and compute need to remain inside your environment.
Your infrastructure, one workflow
Use the same Oumi workflows for data synthesis, evaluation, training, and inference across private environments.
Your team
training data
prod logs
model weights
Evaluate
evals + judges
Synthesize
training data
Train & Optimize
open-weight models
Deploy
serve, scale to zero
Monitor
production traffic
YOUR KEYS
access controls
GPU clusters
training + inference
Analytics storage
Oumi
releases only
ENTERPRISE SECURITY
Secure access to your models, data, and AI workflows.
Oumi provides enterprise-grade security, access controls, and infrastructure options for teams building and operating models and agents across development and production environments.
Role-based access control (RBAC)
Assign organization-level roles and permissions to control who can access projects and perform actions across Oumi.
SOC 2 Type II & SOC 3
Oumi is SOC 2 Type II compliant and has a SOC 3 report covering our security controls.
HIPAA
Oumi supports HIPAA-compliant workloads for organizations handling protected health information.
- RBAC
- SOC 2 Type II
- HIPAA
DEVELOPER TOOLS
Use Oumi from the interface that fits your workflow.
Run model and agent workflows from the command line, integrate Oumi directly into Python applications, or work with Oumi from the AI coding tools you already use.
Oumi CLI
Train, evaluate, synthesize data, run inference, launch jobs, and deploy models directly from your terminal with the Oumi CLI.
Python API
Use Oumi programmatically to build custom pipelines, configure experiments, run workflows from notebooks, and integrate Oumi into your existing ML stack.
Use our MCP with Claude Code and Codex
Work with Oumi directly from Claude Code or OpenAI coding tools. Configure and launch data synthesis, evaluation, training, and other Oumi workflows without leaving your coding environment.
- CLI
- Python API
- MCP
OPEN SOURCE STACK
The open-source foundation behind the Oumi platform.
Oumi OSS is the backbone of our enterprise platform, powering data synthesis, evaluation, training, agentic workflows, inference, and deployment.
Free to use under Apache 2.0, Oumi has earned 9.4K+ GitHub stars and can run on your infrastructure: from local GPUs to cloud and HPC environments.
Anything you build using the Oumi platform is compatible with the Oumi OSS providing you unmatched AI sovereignty.
- Open source
- Apache 2.0
- 9.4K+ GitHub stars
- Run anywhere
Build AI your business owns.
Start with the task. Oumi connects the work, your team controls the decisions, and the intelligence runs where your business needs it.