This website is under construction — information may not be accurate yet. Please use the contact form for more information.
✦ Now Open Source

Deploy Your AI Stack
in Minutes

A full stack AI infrastructure platform covering everything from bare-metal provisioning to GPU inference. All it takes is one command to get started.

$ curl https://get.fairport.io | sudo bash -
Get Started Free Talk to an Expert
One Command
Deploy the full stack
NVIDIA + AMD
GPU support built in
Fully Offline
Works in any environment
Any Model
OpenAI-compatible API
The Platform

From hardware to AI platform,
in one command

Fairport sits between your existing infrastructure and your team. You bring the hardware. Fairport handles everything else.

Any Hardware

Bring what you have

  • NVIDIA & AMD GPUs
  • Bare metal servers
  • Cloud instances
  • Edge & co-lo nodes

One command. Fully configured.

$ curl https://get.fairport.io | sudo bash -
  • Complete stack, zero config
  • GPU to API in minutes
  • Runs fully air-gapped
  • Self-hosted or managed
AI for Your Team

Ready to build on

  • OpenAI-compatible API
  • Web chat interface
  • Access controls & usage
  • Any model, any scale
What's Included

Everything you need.
Nothing you don't.

Fairport wraps best-in-class open-source tools into a single Helm chart. Simple enough for one person to manage, powerful enough to scale to thousands of GPUs.

AI Inference

AI Inference

Deploy vLLM, Ollama, and TensorRT-LLM with auto-scaling. Serves a standard OpenAI-compatible API so your existing tools connect without modification.

GPU Support

GPU Support

First-class NVIDIA and AMD GPU operator support. DCGM metrics exporter ships pre-wired for Prometheus.

Bare Metal

Bare Metal Provisioning

Automated PXE boot and OS installation handle themselves. Go from new hardware to a running cluster without manually touching a single machine.

Networking

Networking

Secure, policy-driven networking without managing physical switches.

Observability

Observability

Prometheus metrics, OpenTelemetry tracing, and Victoria Logs out of the box. Pre-built dashboards for cluster health, GPU utilization, and inference throughput.

Job Scheduling

Job Scheduling

Kueue for priority-based GPU job scheduling with gang scheduling support. Full Slurm compatibility via Slinky lets you migrate existing HPC pipelines without rewriting them.

AI Inference

Any model, standard API

Deploy any open-source model and serve it through a standard OpenAI-compatible endpoint. Your existing apps, tools, and API integrations connect without any modification.

Get started
$ curl http://your-cluster/openai/v1/
chat/completions \
-H "Content-Type: application/json" \
-d '{"model": "llama-3.1-8b",
"messages": [{"role": "user",
"content": "Summarize this report"}]}'
 
{
"id": "chatcmpl-a8f3k2",
"model": "llama-3.1-8b",
"choices": [{
"message": {
"role": "assistant",
"content": "The report covers Q3 results..."
}}
}]
}
Bare Metal Provisioning

From new hardware to running node

Our provisioning layer automates PXE booting and OS installation across your bare metal fleet. Rack a server, plug it in, and Fairport provisions it automatically. No keyboard, no manual setup, no drift.

Live Provisioning
1

Rack your hardware

Active

Connect bare metal servers to the network. No OS installation required.

2

PXE boot & OS install

Active

Fairport detects new nodes, serves a boot image, and installs your chosen OS automatically.

3

Node joins the cluster

Active

The node registers with RKE2, GPU operators install, and it's ready for workloads.

4

Scale to thousands

Active

Repeat for every node. The same process works for one machine or an entire datacenter.

Observability

See everything, from day one

Prometheus metrics, OpenTelemetry tracing, and Victoria Logs ship pre-configured. GPU utilization, inference throughput, and cluster health are visible the moment you deploy. No dashboards to wire up yourself.

87%
GPU Utilization
↑ 12% vs last hour
142
Inference req/s
↑ stable
3
Active Models
↑ all healthy
99.9%
Cluster Uptime
↓ 1 node rebooting
Why Fairport

The hard parts, already handled.

Standing up a production AI cluster from scratch means stitching together a dozen tools, configuring each one, and maintaining all of it. Fairport does that work once so you never have to.

✕

Without Fairport

  • ✕Weeks of manual cluster setup and configuration
  • ✕Each GPU vendor requires separate operator setup
  • ✕Build and maintain your own observability stack
  • ✕Configure a job scheduler separately from your workloads
  • ✕Wire up CNI, ingress, and load balancing by hand
  • ✕Air-gap deployments require custom workarounds
  • ✕No unified inference API across models and hardware
✓

With Fairport

  • ✓Cluster running in minutes with one command
  • ✓NVIDIA and AMD GPU operators pre-configured
  • ✓Prometheus, OpenTelemetry, and dashboards included
  • ✓Kueue and Slurm compatibility built in
  • ✓Networking, ingress, and load balancing wired automatically
  • ✓Air-gap mode ships as a first-class deployment option
  • ✓OpenAI-compatible AI gateway pre-wired and ready
Use Cases

Built for teams like yours

Whether you're running internal AI tooling, building a GPU cloud product, or migrating from HPC, Fairport fits the way you work.

For Engineering Teams

Private AI Cluster

Run your own models on your own hardware. Full control over data, costs, and model selection without depending on external API providers or usage-based billing.

For Hosting Providers

GPU Cloud Offering

Turn raw GPU capacity into a complete AI inference product. Fairport gives you the software stack to compete with major cloud providers without building it from scratch.

For Research & HPC Teams

HPC Migration

Bring existing Slurm pipelines to Kubernetes without rewriting them. Slinky provides full Slurm compatibility so your team keeps working while you modernize the stack underneath.

Getting Started

Up and running in three steps

No Kubernetes experience required. The installer handles RKE2, networking, and storage so you can focus on running models, not maintaining infrastructure.

1

Install the cluster

One command bootstraps RKE2, Cilium, and all prerequisites on your bare-metal or cloud nodes.

$ curl https://get.fairport.io | sudo bash -
2

Deploy the platform

A single Helm release installs GPU operators, the inference stack, monitoring, job scheduling, and more.

$ helm upgrade --install fairport \
  oci://gcr.io/fairport-io/fairport
3

Start building

Deploy models, submit training jobs, configure users, and connect your apps via the OpenAI-compatible API.

$ curl http://your-cluster/openai/v1/
  chat/completions
Pricing

Free to deploy.
Expert help when you need it.

The full platform is free and open source. When you need hands-on support wiring it to your specific environment, our team is here.

Open Source

Self-Hosted

Free forever

Everything you need to deploy a production-ready infrastructure stack on your own hardware, your way.

  • ✓ Full platform via one-line install
  • ✓ NVIDIA & AMD GPU support
  • ✓ Inference, training & monitoring
  • ✓ Slurm + Kubernetes scheduling
  • ✓ Air-gap deployment support
  • ✓ Community documentation & GitHub
View on GitHub →
Expert Support

Managed with Us

Custom pricing

We deploy and tune Fairport for your specific hardware, use case, and scale. Founding engineers, not account managers.

  • ✓ Architecture design & review
  • ✓ Hands-on cluster deployment
  • ✓ GPU hardware procurement & validation
  • ✓ Model selection & optimization
  • ✓ Ongoing managed services & SLA
  • ✓ Direct access to founding engineers
Talk to an Expert →
Solutions

What we help you build

From developers cutting inference costs to datacenters competing with hyperscalers, our advisory services span the full stack.

For Consumers

Developer Advisory

Cost and performance optimization for AI workloads. Professional sourcing of models, providers, and regional capacity. Move beyond expensive APIs with hybrid dedicated/serverless architectures tailored to your SLOs.

For Consumers

Enterprise AI Agents

Transform business processes through intelligent automation. From IT monitoring and incident response to custom workflow automation that delivers measurable operational improvements, all built on your own infrastructure.

For Providers

Hosting Provider Enablement

Level up datacenter operators to compete with major cloud providers. NVIDIA-certified infrastructure, proper cooling design, HA networking, security frameworks, and compliance guidance to accelerate your time to market.

For Providers

Hardware VAR Services

Independent benchmarking and competitive analysis for AI hardware vendors. Installation, integration, and managed services for datacenter providers. We help new hardware find its ideal workload and make the case clearly.

Our Team

Built by people who've done this before

Fairport was founded by infrastructure engineers who helped design and build some of the most demanding AI and cloud platforms in the industry.

NVIDIA GPU Cloud NVIDIA GeForce NOW Palo Alto Networks Wildfire Conviva Performance Analytics SambaNova Cloud

Ready to deploy?

Start for free with the open-source platform, or reach out to have our team handle the deployment for you.

Or email us directly at info@fairport.io