Data Insights and Visualization Engine

Context

Dashboards sit at the centre of how organizations work, yet building them stays locked with engineering teams, and template tools like Power BI trade away the flexibility domain experts actually need. DIVE removes that bottleneck: it lets a non-engineer build a fully customized dashboard just by describing it. You connect to a knowledge graph, select and filter the data in plain language — say, CO₂ emissions for Germany and France from 2014 to 2023 — and generate panels (tables, charts, and forms) by interacting with the engine, refining each one conversationally until it is right. Finished panels flow into a shared, reusable catalogue and assemble into a live, deployed dashboard. The architecture is hybrid neuro-symbolic with a retrieval-augmented generation core: the symbolic side translates intent into precise queries and feeds the model the exact coding context it needs, so the generated interface components come out accurate and consistent rather than improvised. I designed and evaluated it with the people it is for — purchasing and sustainability specialists alongside domain experts — and across every experience level they completed analytical tasks they would normally hand to engineering, found it highly usable with low effort, and pointed to real time and cost savings. The deeper contribution is data democratization: people who cannot write a query get custom, context-aware visualizations without touching code, and the same approach extends naturally from knowledge graphs to relational databases.