Personal project · Analytics product
Trust the path
to the insight.
MetricForge is a local-first BI workbench designed to make dashboards explainable, keep source files in the browser, and bring data quality into the workflow.
The question
A dashboard can look convincing without making its calculations or limitations easy to inspect. I wanted to explore a different starting point: make the data, the logic, and the quality checks part of the same experience.
The approach
MetricForge combines CSV and Parquet profiling with validated DuckDB SQL generation. The local-first design keeps source files in the browser, while data-quality warnings and calculation provenance help users understand what sits behind a visual.
I applied my experience in analytics, KPI design, and reporting to product planning, testing, and AI-assisted development.
What it delivers
- Bar, line, scatter, KPI, and table visuals for different analytical questions.
- Data-quality warnings and calculation provenance to make limitations visible.
- Dashboard persistence and cross-filtering for continued exploration.
- Safe CSV export for taking results into another workflow.
Project scope
MetricForge is a personal project with public source code. It is separate from my professional hospital reporting work.
What matters
Trust grows when people can inspect the route from source data to the number on the screen.
Keep exploring