Building dashboards today requires SQL, charting libraries, and technical expertise. Users who have data and questions but no coding skills are stuck—they need developers to translate their insights into visualizations. A team building analytics tools wanted to collapse that friction entirely by letting anyone describe a dashboard and have AI generate it.
Problem
Data analysts and business users spend weeks waiting for developer time to build custom dashboards. A simple "show me revenue by region over the last three months" requires scoping, development, and iteration cycles. Templates help, but custom analysis remains stuck in a developer queue. The gap between "I have a question" and "I can see the answer" stays wide.
Solution architecture
Natural language to JSON-Render conversion
The platform captures dashboard descriptions via plain-language prompts. A fine-tuned LLM (Gemini) translates those descriptions into json-render specifications—a declarative JSON format that describes UI components, data bindings, and interactivity. No code required.
Semantic data integration
Users upload CSV, Excel, PDF, or JSON datasets. The system auto-detects schemas, infers column types, and stores data in a serverless database. The AI can reference uploaded datasets when generating specs, creating data-bound cards that automatically aggregate and filter.
Draggable canvas with real-time reactivity
Cards are positioned on an infinite canvas with drag-and-drop positioning, resizing, and snapping grids. Cards can be linked with state bindings—selecting a region in one card filters all downstream cards automatically. All updates sync in real-time across connected clients.
Time to first dashboard
< 5 minutes
Components supported
12+ built-in types
Data ingestion
CSV, JSON, Excel, PDF
Card interactions
State-driven filtering
Flexible UI component library
The platform ships with a curated set of components: Metric cards, BarChart, LineChart, PieChart, Table, Form inputs, and Stacks (flexible layouts). All components are declarative, schema-validated, and compose together with state bindings for cross-card interactivity.
Insight
Schema-first architecture
Every component prop is validated against Zod schemas before rendering, preventing runtime errors and enabling AI to generate syntactically correct specs on the first try.
Outcome
Non-technical users can now build custom dashboards in minutes, not weeks. The barrier between "I have data" and "I can see insights" has collapsed. Teams iterate on their own analysis instead of queuing developer time. The AI learns from a corpus of best practices, so generated dashboards are both correct and well-designed by default.
