Interactive portfolio simulation. Responses and processing states are pre-scripted; no file is uploaded and no AI service is called.
VLM / RAG Multimodal intelligence laboratory
Ready to index

Case 04 · Interactive intelligence

Evidence in.
Decisions out.

A multimodal RAG system that reads reports, sees inspection images, retrieves the right evidence, and explains every answer.

03 evidence sources 100% grounded responses LOCAL private by default
INGEST
RETRIEVE
GROUND
CORE ONLINE DRAG TO ROTATE QWEN3-VL / FAISS

Demo responses are selected from a fixed synthetic dataset.

Retrieved evidence

Synthetic source

Portfolio transparency

How this static demo works

This page recreates the interaction design of the working Python application using only HTML, CSS, and JavaScript, so it can run on GitHub Pages.