A leading business school supports researchers working with large, often sensitive datasets across finance, economics, healthcare, marketing, public policy, and organizational behavior. Its researchers wanted AI coding assistants such as Claude Code and Codex for software development and data analysis, without loosening control of that data.
Institutional policy required sensitive research data to remain within the organization's AWS environment, preventing the use of externally hosted AI services. The institution also needed centralized governance over AI model access and detailed visibility into Amazon Bedrock usage costs to accurately allocate AI spending by researcher, project, and workspace.
Relevance Lab integrated Amazon Bedrock with Research Gateway, giving researchers enterprise AI coding assistants inside the institution's own AWS environment, with governance and cost accountability built in.
CASE STUDY • PDF
Secure AI Coding Blueprint
A deep dive into how one business school gave researchers self-service access to Claude Code and Codex on Amazon Bedrock, with every workload and every inference request inside its own AWS environment.
View the Full Case Study →
Business Challenges
The institution wanted to make modern AI coding assistants available to researchers while ensuring institutional control over research data and cloud costs.
01
Secure AI-Assisted Development
Researchers wanted to use AI coding assistants such as Claude Code and Codex to accelerate software development, statistical analysis, and data engineering. However, confidential research data could not be exposed to AI services operating outside the university's AWS environment.
02
Cost Attribution
Unlike traditional infrastructure costs, generative AI usage introduces variable API charges based on token consumption. The institution required detailed visibility into Amazon Bedrock usage costs so that AI expenses could be attributed to individual researchers, research workspaces, and projects alongside traditional infrastructure costs, enabling more effective generative AI cost management.
03
Governed AI Model Access
Research IT needed centralized control over which Amazon Bedrock foundation models, including Claude models, were available to researchers.
04
Self-Service AI Environments
Researchers needed GPU-enabled workspaces on demand, preconfigured and ready to use, without waiting on manual provisioning from Research IT and without the environment drifting from institutional policy.
Solution
RelevanceLab deployed a customized Research Gateway platform accessible exclusively to authenticated university users.
Each research project was provisioned in its own dedicated AWS account, providing strong isolation, centralized governance, and project-level cost management.
Research Gateway was extended with secure AI workspace offerings including:
- Claude Code with Amazon Bedrock
- Codex with Amazon Bedrock
Researchers provisioned AI workspaces through a self-service portal without requiring administrator support.
Each workspace delivered a fully configured AI development environment, ready for immediate use.
- Browser-based remote desktop using Amazon DCV
- CPU and GPU-enabled Amazon EC2 instances
- Pre-installed NVIDIA GPU drivers
- Claude Code pre-installed
- Python, R, Conda, Git, and research tools
- Persistent Research Gateway Study folders
- Secure institutional authentication
Every workspace securely connected to Amazon Bedrock through IAM-based access controls.
- Claude models approved by Research IT
- IAM policies scoping model access per project
- Inference requests kept in-region
- Full audit trail of model invocations
All inference requests remained within the university's AWS environment, ensuring secure and governed access to foundation models.
Research Gateway unified infrastructure and AI usage reporting in a single dashboard.
- Bedrock API costs by researcher
- Bedrock API costs by workspace
- Bedrock API costs by project
- GPU and infrastructure costs
- Total workspace operating costs
Administrators gained complete visibility into the cost of AI-assisted research across projects.
Architecture Highlights
Researchers authenticate to the private Research Gateway portal and provision AI-enabled workspaces through approved catalog offerings. Each workspace runs within its own project AWS account, providing strong isolation between research programs.
Private institutional portal
Claude Code & Codex integration
Study folder data management
Dedicated AWS account per research project
Amazon Bedrock foundation model access
Integrated Bedrock API cost attribution
Self-service AI workstation provisioning
Centralized AI model governance
Project-level cost reporting
Browser-based Amazon DCV remote desktop
IAM-based least-privilege permissions
GPU-enabled compute options
Business Outcomes
Researcher Productivity
- AI-ready workspaces provisioned on demand, not through a Research IT ticket
- Consistent environments across research teams, so setup no longer varies by researcher
- GPU compute available immediately for development and analysis
Secure Enterprise AI Adoption
- All inference requests stayed inside the institution's AWS environment
- Research IT controlled which Bedrock models researchers could reach
- Each project isolated in its own AWS account under least-privilege IAM
Operational Efficiency
- Self-service provisioning removed manual builds from the Research IT queue
- Standardized workstation images maintained centrally, updated once for everyone
- GPU infrastructure managed through the catalog rather than per request
Financial Visibility
- Bedrock token spend attributed to individual researchers, workspaces and projects
- AI and infrastructure costs reported together in one dashboard
- Budget forecasting grounded in actual usage rather than estimates
Customer Perspective
“
Research Gateway enabled the Business School to adopt AI-assisted software development in a secure and governed manner. By combining Claude Code and Codex with Amazon Bedrock, our researchers can leverage state-of-the-art coding assistants while keeping sensitive research data within the organization's AWS environment. The platform has simplified onboarding, improved visibility into AI usage costs, and significantly reduced operational overhead for Research IT.
— Research IT Leadership
Research Gateway: Built for Modern AI Research
Research Gateway is Relevance Lab's cloud-native platform for secure research computing, AI-enabled environments, and scientific collaboration. It helps universities, healthcare organizations, government agencies, and research institutions deploy governed cloud workspaces, integrate enterprise AI services like Amazon Bedrock, and give researchers self-service access to compute and AI tools, all while keeping costs visible and meeting institutional security and compliance requirements.
We combine deep AWS platform expertise with hands-on experience in research computing, helping institutions standardize AI-ready workspaces, reduce day-to-day dependency on Research IT, and accelerate discovery without compromising governance.
AI-ENABLED RESEARCH CASE STUDY
Ready to Give Researchers Governed Access to Claude?
Relevance Lab builds AI-enabled research platforms on AWS, from Bedrock model governance to per-project cost attribution. See how it worked for one business school.
