Inspiration
What it does
How we built it
Challenges we ran into
Accomplishments that we're proud of
What we learned
What's next for Veritas Nexus
Veritas Nexus
Inspiration
The rapid growth of AI-generated content and the increasing number of paper retractions have made research integrity a growing concern. Academic reviewers and researchers often need to manually verify hundreds of citations, making it easy for fake, unverifiable, or retracted references to slip through the review process.
We wanted to build an autonomous system that could act as a "fact-checker for academic citations" and help researchers, universities, publishers, and reviewers quickly assess the trustworthiness of a paper.
What it does
Veritas Nexus is an autonomous research integrity platform that audits academic papers and generates a trust score based on citation quality.
Given a paper, the system:
- Extracts and analyzes citations
- Verifies references against Crossref metadata
- Detects fake or unverifiable DOIs
- Flags citations linked to retracted research
- Computes an integrity score
- Generates a detailed audit report
- Publishes findings to a searchable knowledge base
- Sends alerts and observability metrics automatically
In our latest test run, Veritas Nexus analyzed a paper containing 260 citations, detected 11 unverifiable references, identified 12 retracted citations, and generated an integrity score of 96.5/100.
How we built it
We designed Veritas Nexus as a multi-agent pipeline integrating several sponsor technologies:
Data Ingestion
- Airbyte synchronizes citation metadata into ClickHouse.
- Crossref data is used as the source of truth for citation verification.
Verification Engine
- Python agents validate citation metadata.
- DOI records are cross-checked for existence and consistency.
- Retracted references are identified and flagged.
Integrity Scoring
The platform computes an integrity score based on citation quality, retractions, and verification results.
Reporting
- Detailed Markdown audit reports are generated automatically.
- Reports are published through Senso for downstream consumption.
Alerting & Monitoring
- Composio-powered alerting workflows notify users of integrity issues.
- Langfuse captures scores and observability data for every run.
Dashboard
- An OpenUI-based dashboard provides real-time visibility into citation quality, integrity scores, and verification results.
Challenges we ran into
One of the biggest challenges was integrating multiple sponsor technologies into a single autonomous workflow while maintaining reliability.
We also encountered authentication challenges when configuring Composio OAuth due to institutional Google account restrictions. To keep the pipeline operational, we implemented a fallback alert logging mechanism while preserving the alerting architecture.
Another challenge was handling inconsistent citation metadata and distinguishing between missing records, malformed references, and truly retracted works.
Accomplishments that we're proud of
- Built a fully autonomous end-to-end research integrity pipeline
- Integrated Airbyte, ClickHouse, Senso, Langfuse, Composio, and OpenUI
- Successfully verified hundreds of citations in real-world papers
- Automated reporting, scoring, alerting, and observability
- Achieved one-command execution of the entire workflow
- Implemented secure secret management using environment variables
What we learned
This project taught us how difficult research verification remains despite the availability of large academic databases. We learned how to build reliable agent-based workflows, integrate observability into AI systems, and design data pipelines that can automatically detect integrity issues at scale.
We also gained valuable experience working with sponsor technologies and connecting them into a cohesive autonomous system.
What's next for Veritas Nexus
- Expand ingestion to larger academic datasets
- Add AI-powered claim verification beyond citation checking
- Support publisher and journal workflows
- Enable real-time monitoring for newly published papers
- Enhance observability with full Langfuse tracing
- Deploy automated email and collaboration workflows through Composio
Our long-term vision is to create an automated integrity layer for scientific research that helps prevent misinformation, citation fraud, and the propagation of retracted science.
Log in or sign up for Devpost to join the conversation.