ATLAS – Scientific Evidence Workspace

Inspiration

Researchers spend a significant amount of time collecting, validating, and organizing scientific literature before any real analysis begins. ATLAS was created to simplify that process by providing a dedicated workspace for evidence management.

What it does

ATLAS allows researchers to:

• Import studies using DOI or PMID.
• Validate bibliographic metadata.
• Detect duplicate records automatically.
• Organize a structured evidence library.
• Prepare datasets for downstream AI-assisted evidence synthesis.

How we built it

ATLAS was built using GPT-5.6 and Codex during OpenAI Build Week.

The application uses a modern React/Next.js frontend with TypeScript and OpenAI Sites. Persistent storage is handled through Cloudflare D1 and document storage through Cloudflare R2.

Challenges

The biggest challenge was designing a reliable workflow for scientific evidence management while maintaining accurate metadata validation and duplicate detection. Another challenge was integrating multiple services into a simple user experience suitable for research teams.

Accomplishments

The current prototype successfully supports:

• DOI/PMID import
• Metadata validation
• Duplicate detection
• Persistent storage
• Searchable evidence library

What we learned

Building with Codex dramatically accelerated development and enabled rapid iteration from idea to working prototype during Build Week.

What's next

Future releases will include:

• AI-powered evidence extraction
• Full-text document parsing
• PRISMA workflow support
• Collaborative annotation
• Systematic review automation

Built With

• cloudflare
• codex
• drizzle
• gpt-5.6
• next.js
• openai
• react
• sqlite
• typescript