Skip to content

AI Factories

Two factories. One builds your AI. One runs it.

$ megalab factory --list
1  cloud   builds it   NVIDIA Blackwell GPU cluster
2  edge    runs it     on-site, offline-ready, your hardware

1 · Cloud AI Factory

Builds it. Our NVIDIA Blackwell GPU cluster for the heavy lifting.

  • train

    Train and fine-tune models on your own data.

  • scale

    Big-model inference and agents, at scale.

  • burst

    Extra capacity when you need it, gone when you don't.

  • zero-hardware

    No GPUs to buy, rack, or babysit.

2 · Edge AI Factory

Runs it. Your compute lives with you, not in someone else's cloud.

A three-floor edge compute factory, tied together east-west by a Reticulum-powered WAN fabric. Workloads move, scale, and fail over without a central cloud.

  • floor 1 · CPU

    The application floor. Web apps, APIs, dashboards, databases, queues, automation.

  • floor 2 · GPU

    The AI workshop. Models from the Cloud AI Factory, quantized and compressed to run on your own nodes. No hidden cloud bill.

  • floor 3 · Edge

    The field floor. CPU + GPU on your local network, talking to your hardware. Data stays on-prem.

how they work together

  1. Train big in the Cloud AI Factory, on your data
  2. Quantize the model down to edge size
  3. Deploy it to the Edge AI Factory, on your site
  4. Own it. It keeps running when the internet doesn't

The difference: the Cloud factory builds your models. The Edge factory runs them.

shared clusters

Public and private APIs · MCP services · inference · training · agents · vector storage · automation