Uh, hi, everyone. Uh, in this demo, we are going to demonstrate Armada Bridge, uh, platform as a service features. I'm now first logging in as, uh, the administrator. And these are admin dashboard, where the admin can monitor all the, uh, resources here, free GPUs, how many VMs and bare metal servers has been created. And then we go into the catalog. Uh, in the catalog, the admin can see, uh, the server, uh, availability and the SKUs, and similarly for virtual machines. And we have one already onboarded a tenant. In the view details, uh, administrator can set the resource quota by configuring- by clicking configure quotas, and similarly, set pricing by clicking configure a resource price. We'll now go inside the tenant, and these are tenant dashboard where tenant can monitor other, uh, available resources and the usage. Under compute section, uh, when we click on Virtual Machine, um, we have already created few virtual machines, and, uh, the communities clustered with required templates has alread- al- has also been created. A VM overview, and we also have CPU and GPU observability tabs for the virtual machine. And these are cluster page, uh, we have already created, uh, one cluster with, uh, GPU access with model serving template, and another cluster with, uh, Jupyter Notebook plus SKY-scheduler as a template. The admin-armed users can also get access to the communities clustered terminal within the UI. And when we describe the node, uh, the GPU is attached to it. Here's 200 GPU with MIC profiles. There are, uh, two 1G profile and two one- uh, 2G profile. Uh, we will now reconfigure, uh, the MICs from the bridge platform by going to node and click on the node, and edit the GPU profile. So we will remove, uh, 2G profiles and we'll add one more 3G profile, and click on Update. So the platform will dynamically reconfigure the br- uh, the MIC profiles. We will confirm that by executing NVIDIA SMI- ... on the virtual machine. Now we see one 3G profile and two 1G profiles configured. And similarly, we can go to cluster terminal here, and we can also verify. And, uh, now the admin can click on the AI models page where, uh, the admin can see that, uh, deployed models and also the Jupyter Hub. And for Jupyter Hub, uh, just click on Add New Server, and it will pop up a window to give a name to the, uh, environment, and select the curated images, and either with the GPU with- environment with the GPU access or with GPU MIC profile. So we'll select one of the MIC profile and start the server. Now the isolated, uh, development environment notebook has been spun up for the user, so the user will just click on, uh, the URL here, and it will open a Jupyter Notebook environment, and this environment can be used for all the development, uh, work- development work for that user. Right, I can c- I can run any Python, uh, commands, and, uh, this, this, uh, environment is now attached with the GPU MIC profile that can be used for AI/ML workload development-related tasks. So now we see the SLURM. So it's a managed, uh, SLURM cluster. So we have already created a managed clum- SLURM cluster with one master and one worker node, and, uh, these are dashboard for the SLURM cluster. Two nodes and, uh, 84 cores and 208 memory. And similarly, the resource page will show the faster resources. And in the jobs page, uh, the user can submit, uh, the machine learning Hype- HPC-related kind of, uh, jobs through this page here. Right, uh, now we see, uh, the model as a service. So we also have created a cluster with, uh, (inaudible 06:34) as a template, and a Qwen model is also deployed. And, uh, the endpoint and the model is also created, so we can see the m- deployed model in the deployed models page here. So we have deployed Qwen 2.5 billion model. We go to the models page and we see the deployed models here. Uh, click on new details to the- to see the pricing and, uh, the curl endpoint to interacting with the model programmatically, and we also have a endpoint here which can be integrated with, uh, external chatbots or AI agents. So in this demo, uh, we just click on Open in Playground and interact with the model and see how we- how the user can get the real-time response. So before that, uh, let's create a API key for authentication. And query the model and see the response from the model in real time. And based on the token's usage, uh, the platform will provide the usage metrics for the administrator, uh, to charge back the user based on the tokens. Thank you for watching

Armada Bridge Platform User Guide