Case Study - Cross-Repository Retrieval Infrastructure with Enforced Access Boundaries

System Architecture Snapshot

• Data Layer — Box, Google Drive, SharePoint connectors; document chunking and embedding
• Retrieval Layer — RAG pipeline with Gemini (Vertex AI) for long-context analysis
• Control Layer — Authentication, programmatic user provisioning, source attribution enforcement
• Interface Layer — Conversational query interface with chat history continuity

The Challenge

Palm Health Foundation — a philanthropic organization funding mini grants for individuals, small nonprofits, and small businesses — had years of historical mini grant data scattered across multiple storage platforms: Box, Google Drive, and OneDrive/SharePoint.

This included Excel tracking spreadsheets, instructional documents, templates, interview summaries with project directors, and physical sign-in sheets.

The foundation wanted to identify themes and patterns across years of grant activities to inform future funding decisions. But with data siloed across three platforms and stored in varied formats, there was no efficient way to analyze it holistically.

The Solution

BeeNex engineered an internal retrieval application that connects to all three data repositories and enables foundation staff to ask natural language questions about their historical grant data.

The system:

• Connects to Box, Google Drive, and OneDrive/SharePoint to ingest and index grant documents
• Uses Gemini (Vertex AI) for long-context understanding, chosen for its speed and cost-effectiveness with large document sets
• Provides a conversational interface where staff ask questions like "What themes emerged across 2022 mini grants?" or "Which grantees focused on youth mental health?"
• Maintains chat history so users can pick up and continue conversations
• Includes authentication with programmatic user creation for foundation staff

Key Deliverables

• Private AI chat web application deployed to GCP
• Multi-source data connectors (Box, Google Drive, SharePoint)
• RAG pipeline for document retrieval and analysis
• Authentication system with email-based registration
• Chat history and conversation continuity
• Onboarding guide, staff training, and walkthrough