Saikat Kumar

We built a secure RAG engine & lightweight DMS. What features are we missing?

by

Hey everyone,

Our small engineering team just finished building the core architecture for CordonData, a secure RAG engine and lightweight Document Management System (DMS) built for enterprise AI.

Rather than talking about how complex data pipelines are, I want to share exactly what we’ve built to solve the infrastructure side of AI, and get your direct feedback on what we should build next.

Here is what the platform currently handles out-of-the-box:

  • Cross-Silo Sync via Standard Protocols: We natively connect to your existing organizational silos using standard enterprise protocols like CMIS, REST APIs, and FTP.

  • Markdown Extraction: Instead of dealing with heavy raw files, we extract and synchronize the content from those external silos into clean, structured, LLM-ready Markdown.

  • Hybrid Search: We run a dual-engine approach combining traditional BM25 keyword ranking with semantic vector embeddings to ensure highly relevant context retrieval.

  • Zero-Trust Security & Universal Audit: We natively enforce Document-Level Security (DLS) by syncing your existing access controls. The AI only retrieves data from files a specific user is explicitly authorized to view. Every query and document access point is meticulously audited and logged for complete enterprise traceability.

  • Built-in DMS: For teams without a centralized hub, we provide a clean, lightweight DMS interface to upload and manage files directly.

We are currently bringing on a few early Design Partners for white-glove onboarding. But before we lock in the final roadmap for our public launch, we’d love your validation:

  1. The Feature Gap: Based on the core engine described above, what is the #1 feature, dashboard, or capability you would need to see added before you would deploy this for your own users?

  2. Architecture Validation: Does the combination of a hybrid search engine tightly coupled with a lightweight DMS fit into your current workflow, or do you typically prefer those layers completely separated?

Any suggestion or feedback is most welcome, Would love to hear what the builders here think!

25 views

Add a comment

Replies

Be the first to comment