This guide gets your cloud environment working before the bootcamp so we can spend our time on AI and biomedical computing instead of setup. It walks you through logging into GW Research Gateway (AWS), launching a Rocky Linux 8 Remote Desktop, starting Jupyter, and pulling the model with Ollama.
Why AWS and not Pegasus? Pegasus's GPFS storage is unavailable for the bootcamp window. Research Gateway gives everyone a full Linux desktop today, and every skill we use — Jupyter, PyTorch, Ollama, bash, Python environments — transfers directly. The main thing that does not carry over is Slurm, which we do not need for this bootcamp.
Plan for about 30 minutes. The instance itself takes ~15 minutes to build, so start early.
By the end you should have all of these done:
- Signed in to Research Gateway at rg.arc.gwu.edu
- Launched a Rocky Linux 8 Remote Desktop (80 GB storage,
t3.2xlarge) - Logged into the desktop and opened a terminal
- Started Jupyter from a Python virtual environment
- Opened and run
ollama_aws.ipynb, including pulling the model
Come to the bootcamp with all five checked.
README.md— the pre-bootcamp setup guideollama_aws.ipynb— the main notebook for Research Gateway / AWSollama_colab.ipynb— the Colab version of the same exerciseimages/— screenshots used in the setup walkthrough
- Your first login. The very first time you sign in, the portal shows a "requires admin approval" message. This is normal — but until an admin approves you, you can't launch anything. Logging in early means approval is already granted when you need it.
- Confirm you're on the
AI-ML_BioMedproject. If you don't see it after signing in, contact the organizers before the session. - Use a current browser (Chrome works well). You'll end up with two tabs: the portal and your remote desktop.
- Go to https://rg.arc.gwu.edu/
- Click InCommon-Login.
- On the next screen, choose Sign in with GW and authenticate with your GW credentials.
First login? You'll see a message that your account requires admin approval. That's expected. Once it's granted you'll land on the projects screen.
From My Projects, open the AI-ML_BioMed project card.
Go to the Available Products tab and select Rocky Linux 8 Remote Desktop.
⚠️ There are look-alikes in the list — Rocky Linux 8 EC2 and Windows Remote Desktop. Make sure you pick Rocky Linux 8 Remote Desktop (circled below).
Give the product a name (letters, numbers, dots, hyphens, and underscores only — no spaces), then scroll down and set:
| Setting | Value |
|---|---|
| Product Name | Anything memorable, e.g. MyAWSInstance |
| EBSVolumeSize | 80 (GB) |
| InstanceType | t3.2xlarge |
| Password | Choose your own — you'll use it to log into the desktop |
Then click Launch Now.
⏱ Expected provisioning time is about 15 minutes. You can keep working in the portal or step away while it builds.
🔒 About the password: you set your own password here. Ignore any pre-filled default shown on the product description — pick something only you know.
When the instance is ready, it opens in a new browser tab showing a Rocky Linux lock screen.
- Log in with the password you created in Step 4.
- If you don't see a password box, click inside the window and drag upward to reveal it.
Open a terminal on the desktop and run these commands one at a time:
sudo dnf install python3.12 -y
python3.12 -m venv my_new_env
source my_new_env/bin/activate
pip install notebook
jupyter notebookA Jupyter notebook will open in a browser inside the remote desktop.
Important: the virtual environment must be active (the
source my_new_env/bin/activateline) before you runjupyter notebook. If you open a new terminal later, run thatsourceline again first.
In the Jupyter file browser, open:
/home/ec2-user/studies/ProjectStorage/ollama_aws.ipynb
Run the notebook top to bottom. Run the model-pull cell early — the download takes a while, and running it now means the model is ready before the bootcamp instead of during it. When a later cell gets a response from the model, you're done.
| Problem | Try this |
|---|---|
| Login fails | Re-check the InCommon / Sign in with GW steps. If it's your first login, your account may still be waiting on admin approval. |
| Nothing to launch / no project | Confirm you're a member of the AI-ML_BioMed project and that it shows a budget. Contact the organizers if it's missing. |
| Instance won't launch | Double-check the config: 80 GB storage and t3.2xlarge. Make sure the product name has no spaces or special characters. |
| Can't find the password box | Click in the desktop window and drag upward to reveal it. |
jupyter: command not found |
The virtual environment isn't active. Run source my_new_env/bin/activate, then jupyter notebook again. |
| Package or model missing | Re-run the pip install / model-pull cells with the venv active. |
- RTShelp — for account, access, and system issues.
- Office Hours — bring anything you could not get working before the session.
Please sort out access before the bootcamp so we don't spend session time on setup.
The t3.2xlarge instance bills against the shared $5,000 AI-ML_BioMed project budget the whole time it's running. When you finish for the day, stop your instance from the portal so it isn't quietly burning the budget.
Portal: https://rg.arc.gwu.edu/ → InCommon-Login → Sign in with GW
Project: AI-ML_BioMed
Product: Rocky Linux 8 Remote Desktop
Config: 80 GB storage · t3.2xlarge · your own password
Notebook: /home/ec2-user/studies/ProjectStorage/ollama_aws.ipynb
# On the desktop, in a terminal:
sudo dnf install python3.12 -y
python3.12 -m venv my_new_env
source my_new_env/bin/activate
pip install notebook
jupyter notebook