For Research Teams¶
RosettaScience and RosettaInsight -- reproducible, governed, multi-cloud research computing.
Overview¶
Research teams need powerful compute without the overhead of cloud operations. Most researchers are not cloud engineers -- and they shouldn't have to be. RosettaHub's MetaCloud gives principal investigators and researchers self-service access to HPC, GPU, and general-purpose workloads across AWS, Azure, GCP, Alibaba Cloud, OVH, and OpenStack -- with grant-aligned budgets, reproducible environments, and zero cloud expertise required.
The platform's value goes beyond multi-cloud: it makes cloud computing accessible to a much wider audience. A researcher who has never opened the AWS Console can launch an HPC cluster, snapshot their environment, and share it with collaborators -- all through formations and one-click workflows.
Key Capabilities¶
Reproducible Environments via Formations¶
Formations are cloud-agnostic recipes that capture an entire research environment. The research workflow follows a natural cycle:
Clone a public or shared formation Customize packages and data Launch on any cloud Snapshot your work Share with collaborators Publish to the marketplace Iterate
Because formations are cloud-agnostic, an environment built on AWS deploys identically on Azure or GCP -- enabling cross-institutional collaboration without lock-in.
Grant-Aligned Budget Delegation¶
RosettaHub's cloud operations layer maps directly to how research is funded:
| Level | Maps To | Controls |
|---|---|---|
| Organization | Department or Institute | Top-level budget pool |
| Sub-Organization | Research Group | Delegated budget with transfer rights |
| Project | Grant or Experiment | Dedicated cloud accounts with hard spending limits |
| User | Individual Researcher | Zero, one, or more dedicated cloud accounts with personal budgets -- or participation in a project's cloud accounts via project roles |
Roles can be assigned directly to users or inherited from the organizations they belong to. A user with no personal cloud accounts can still work within a project's cloud accounts through assigned project roles.
Budget enforcement is real-time and event-driven -- not based on billing lag. If a grant allocation is exhausted, new launches are blocked immediately.
Self-Service Portals¶
Researchers access pre-approved formation catalogs without filing IT tickets. Administrators define guardrails (instance types, regions, spending limits) and researchers operate freely within those boundaries.
Data Management and Sharing¶
Formations bundle data and compute together, so researchers receive everything they need in one click -- no separate data configuration step.
| Capability | How It Helps Researchers |
|---|---|
| Formations bundle data + compute | Storage mounts are part of the formation template. Launch an environment and the datasets are already there. |
| Shared storages | Share storages (S3, EFS, EBS) with fine-grained access control. Curated datasets can be shared read-only while each researcher has personal writable space. |
| Cross-cloud data mounting | Mount AWS S3 data on a GCP instance -- cloud boundaries are transparent for data access. |
| Scoped dataset views | Map specific folders within a large S3 bucket to individual MetaCloud storage artifacts, giving teams focused access to the data they need. |
| Private marketplace | Publish formations (with dataset mounts) to institutional catalogs -- effectively a research service and data catalog. |
| URL-based sharing | Share a dataset-configured formation as a link. Collaborators deploy it with data already mounted, on their own budget. |
Multi-Cloud HPC and GPU¶
Launch HPC clusters (AWS ParallelCluster), Spark/Hadoop clusters (EMR, Dataproc), and GPU instances (NVIDIA) from a single interface. Spot and preemptible instances reduce costs by up to 90% for fault-tolerant workloads.
AI and GenAI Integration¶
Researchers can access cloud AI services -- AWS Bedrock, SageMaker, and others -- through federated console access while RosettaOps governs the budget. GPU formations provide compute for model training and inference.
| Capability | Status |
|---|---|
| GPU formations for ML training | Available |
| Federated access to cloud AI services (Bedrock, SageMaker, Vertex AI) | Available |
| AI/LLM usage tracking and budgeting | Coming Q2 2026 |
| MCP server for AI-powered research operations | Coming Q1 2026 |
| ResOps Agents for AI-driven governance | Coming Q2 2026 |
Trusted Research Environments¶
For research involving sensitive data, RosettaHub supports Trusted Research Environments aligned with the Five Safes framework. Project isolation, encrypted storage, SSO, RBAC, and compliance policies provide strong coverage across Safe People, Projects, Settings, and Data. Egress controls (Safe Output) are being strengthened with Gitea airlock and Amazon Macie integrations.
RosettaInsight on AWS Marketplace¶
RosettaInsight is RosettaHub's MetaCloud offering on AWS -- one-click GPU/CPU instances, pre-built template-based environments, and team collaboration tools delivered as a self-service platform. Researchers launch compute, share environments, and work with familiar tools (Jupyter, RStudio, VS Code) without needing AWS expertise.
Free Tier Available
RosettaInsight includes a free tier (up to 4 GB RAM) with Jupyter, RStudio, and VS Code environments -- ideal for evaluating the platform with a small research group. Paid plans start at $10/month per 8 GB RAM.
Next Steps¶
- Research Workflows -- concrete worked examples for research personas
- Quick Start -- connect your first cloud account
- Formations -- learn how to create and share environments
- Trusted Research Environments -- Five Safes alignment for sensitive data
- Tutorials -- step-by-step walkthroughs