Data Quality Scorecard — Trust Your Data

Overview

Data Quality Scorecard evaluates how trustworthy a dataset is before it feeds a report, model, or pipeline. It runs deterministic checks over an in-memory SQLite engine — missing values, duplicate rows, statistical outliers (IQR), constant columns, and inconsistent formats (e.g. mixed date or email patterns) — and produces a per-column and overall quality score with severity-rated issues. An AI reviewer then summarizes the findings and prioritizes concrete fixes. The checks are computed locally; the LLM only summarizes. LLM-agnostic — runs on a local model (Ollama / Llama) or any cloud LLM; this deployment uses Gemini. Data is processed in-memory and expires automatically.

Problem Solved

Bad data silently breaks dashboards and models, and quality problems are usually discovered far too late. This demo gives an immediate, quantified read on data trustworthiness — what is wrong, how severe it is, and what to fix first — turning "data readiness" from a slogan into a measurable score.

Capabilities

• Missing-value, duplicate, and constant-column detection
• Statistical outlier detection (IQR)
• Format-consistency checks (dates, emails, phones, …)
• Per-column and overall quality scoring with severity
• AI-prioritized, actionable recommendations
• CSV upload or built-in sample datasets
• In-memory processing, rate limiting, gated access