Your first session
Sign in
Go to run.meshia.io and sign in with your email. No password. We send a 6-digit code. Check your inbox, enter the code, and you're in.
Pick a GPU
On the dashboard, click New session. You'll see a list of available GPUs with their VRAM, hourly price, and current availability. If you're not sure what to pick:
- RTX 4090 (24 GB): good default for fine-tuning models up to 13B parameters
- A100 (80 GB): large-scale training, big batch sizes, multi-GPU jobs
- H100 (80 GB): fastest option for anything compute-bound
Pick one and hit Select.
Name your session
Give your session a short name like "llama-finetune" or "diffusion-exp-3". This shows up in your dashboard, experiment logs, and agent memory. Keep it descriptive. You'll thank yourself later.
What happens during provisioning
After you click Launch, Meshia provisions a GPU pod. This usually takes 30 to 90 seconds. You'll see a status indicator move through: queued → provisioning → starting → running. The workspace opens automatically once the pod is live.
Your first message
Once the workspace loads, you'll see the chat panel on the right. Type something like:
Set up a Python environment with PyTorch 2.2 and download the Alpaca dataset
The agent will start executing immediately, installing packages, writing scripts, downloading data. You can watch the terminal output in real time in the left panel.
That's it. You're running on a GPU with an AI agent. Everything from here is just telling it what to do next.