MLViz: business-value charts for ML classifiers

A web app that helps teams show the business value of a classification model, beyond accuracy. Users upload scored test data (CSV, XLSX or Parquet) and get lift, gain and profit charts that update as they change cost and revenue inputs.

• File parsing runs in Web Workers with parquet-wasm and Apache Arrow, so large uploads don’t freeze the page
• Custom metrics written in JavaScript in a built-in code editor
• AI chart explanations streamed from Amazon Bedrock (Claude)
• Shareable views on AWS Amplify, plus a Python package that publishes a view straight from a Jupyter or Databricks notebook

Stack: React 19, TypeScript, Mantine, Observable Plot, ECharts, AWS Amplify, Amazon Bedrock, Python