Engineering decisions
- Vector retrieval grounds responses in organization-scoped knowledge.
- Configurable providers reduce coupling to one model API.
- Unresolved questions escalate to HR staff instead of being guessed.
Back to selected work
Case study · 02
AI HR Assistant
An enterprise RAG assistant for organization-scoped document search, evidence retrieval, and configurable AI providers.
- Category
- Enterprise RAG assistant
- Availability
- Production system · Public case study
Problem + approach
Overview
Teams need fast answers from internal documents without losing traceability or depending on a single model provider.
Technology stack
- NestJS
- Qdrant
- PostgreSQL
- OpenAI
- Gemini
- RAG
Conceptual overview
System architecture
- 01Role-based React consoles
- 02NestJS API + role guards
- 03Chat + document services
- 04Qdrant + HRIS
- 05OpenAI / Gemini
- 06Application database
This project-specific overview is based on the published case-study architecture. Internal implementation, credentials, and deployment details remain undisclosed.
Decisions + constraints
Engineering notes
Technical challenges
- Retrieval relevance across varied documents
- Keeping generated answers grounded in source context
Results & impact
- Grounded enterprise document search delivered at Apping Technology
- Human handover for questions the knowledge base cannot answer
Lessons learned
- Useful RAG depends as much on retrieval quality and context control as on the selected LLM.
Screenshots + diagrams
Project media
Availability + roadmap
Documentation
Repository & API documentation
Future improvements
- Retrieval evaluation suite
- Source-level feedback