projectEureka Technical Details

EnvironmentAWS, GCP, or on-prem K8s

BucketsAWS S3 & GCP GCS

AppsJupyter, VS Code, desktop

Vision: project-centric platform

Interactive apps, secure agents, environment placement, and storage, around one project spine.

Interactive Browser launchers and persistent workspaces

Agents Run in a secure environment; access from anywhere

Environment Deploy on AWS, GCP, or on-prem Kubernetes

Storage Home dir, project storage, multi-cloud buckets

Interactive: browser launchers (Jupyter, VS Code, desktop, terminal) and project-focused workspaces that persist across devices.
Agents: run in a secure project environment, keep working when your laptop sleeps, access and supervise from anywhere in the browser.
Environment: deploy on AWS, GCP, or on-prem Kubernetes, one target under your control. Free software for teaching and research; you cover infrastructure costs.
Storage: home directory, project directories, plus multi-cloud buckets (AWS S3, GCP GCS, more to come), with project-based access next to the apps.

Where the environment runs

Deploy projectEureka™ on one environment target: Amazon Web Services, Google Cloud, or on-premises Kubernetes. Attach AWS and GCP buckets to projects regardless of where the environment runs.

AWS or GCP: Run the environment in the cloud account and region you choose.
On-prem Kubernetes: Deploy where policy or data residency requires local control.
Apps inside that environment: Jupyter, VS Code, Linux desktop, and other launchers.
Company capability: Omnibond has delivered cloud HPC at extreme scale (see below); that engineering depth sits behind the platform.

Create a project in projectEureka — Projects live inside your chosen environment

Project storage and multi-cloud buckets

Multi-cloud shows up in storage: home directories, project directories, and multi-cloud buckets you can attach no matter where the environment runs (AWS, GCP, or on-prem K8s).

Home directory: Personal space for the user inside the environment.
Project storage: A project directory that stays with the team workspace.
Multi-cloud buckets: AWS S3 and GCP GCS today, more providers to come, usable from any deploy target.
Project-based access: Permissions follow the project.
OrangeFS® path: Parallel file system when jobs outgrow a single mount or object layout.

Buckets and project storage under project control

Interactive apps and virtual workstations

Launch tools in the browser next to project data. Personal machines remain the client; the environment is a secure, consistent virtual workstation for the project.

Launchers: Jupyter, VS Code Server, Linux desktop, terminal, and related tools.
Persistent workspaces: Long-running work continues when devices sleep or leave campus.
Resource visibility: Cores, RAM, and GPU shown on running apps in the UI.
Shared app model: The same launcher and workspace pattern spans teaching, research, and enterprise (licensing differs).

Desktop and notebook apps in the browser

OrangeFS® storage

Optional scale-out parallel file system for AI and HPC workloads when object storage or a single mount is not enough.

Linux kernel integration: Native client path for high-performance mounts.
Distributed metadata: Scale-out architecture for aggregate throughput.
Multi-platform clients: Linux kernel, Windows client, and container mounting.

Commercial OrangeFS → · Community open source →

Storage resources view — Storage resources in the product UI

User-level security architecture

Open OnDemand–style isolation: shared auth frontend, per-user apps and sessions.

Eureka-UI uses a shared frontend for authentication and reverse proxy, then isolates interactive and HPC apps per user.

Identity: Federation and enterprise IDPs (Shibboleth, Entra, and related paths).
Server frontend: Authentication and reverse proxy (Apache user context).
Per-user NGINX: Runs as the individual user for app isolation.
Per-user JWT: Tokens scoped to the user for scheduler and app paths.

User-level security architecture based on Open OnDemand: identity providers, Apache frontend, per-user NGINX, Eureka-UI, and OmniSched — Based on Open OnDemand

Project-oriented architecture

Today’s model is simple: the project is the spine. People, apps, and storage hang off it inside one environment you deploy.

Environment

One deploy target: AWS, GCP, or on-prem Kubernetes

Project

Name, description, collaborators, and access; the spine everything else hangs off

Apps & agents

Jupyter, VS Code, desktop, terminal; agents run securely and stay reachable from anywhere

Storage

Home directory, project storage, multi-cloud buckets (AWS S3, GCP GCS, more to come)

Teaching and research use free software (you still pay cloud or cluster costs). Enterprise is a separate licensed path on the same platform model.

Direction: building infrastructure for what’s next

projectEureka™ is shipping a clear “now” model. In parallel, Omnibond is investing in platform infrastructure so future capabilities can land on the same project spine. Items below are direction.

Now

One environment target: AWS, GCP, or on-prem Kubernetes
Project spine: people, apps, agents, and access
Multi-cloud buckets: AWS S3 and GCP GCS (more storage providers over time)
Browser apps & virtual workstations next to project data
Secure long-running work in the project environment

Building toward

Job routing: smarter placement of work across queues and capacity pools
SLURM compatibility: familiar batch workflows alongside interactive project apps
Dynamic filesystems per job: mounts and scratch that follow the workload
Additional platforms: more environment targets as the control plane matures, including Azure as future work. Supported deploys today remain AWS, GCP, and on-prem Kubernetes
Richer multi-environment orchestration: company-scale HPC experience applied to future job placement. PE today remains one environment target
Deeper agent workflows: lifecycle, supervision, and policy hooks for agentic AI and simulation jobs
Stronger parallel storage paths: including OrangeFS® and job-adjacent high-throughput I/O
Data staging & lifecycle automation: prepare, place, and retire datasets as jobs and projects evolve