Project Brief

Assessment Intelligence

An interactive decision layer that turns SQL-backed assessment records into stakeholder-ready views of growth, performance, participation, completion, and data quality.

Open Dashboard Read R Report GitHub Repo

Overview

Decision layer for assessment stakeholders

The dashboard helps leaders see whether students are improving, where performance or participation signals are weakest, which groups warrant closer review, and whether the source data is reliable enough to support action. The current DuckDB analytics layer uses five SQL-backed synthetic extracts with 3,020 rows, including 2,009 readiness records across 174 course-section groups.

Inspect First

  • Hosted synthetic assessment dashboard
  • Rendered R gradebook synthesis report and PDF export
  • Validation summary for the synthetic gradebook reconstruction
  • SQL warehouse assessment report in the source repository
  • Dashboard data extracts generated from public-safe synthetic marts
  • Minimum-sample safeguards for ranked decision insights
Dashboard R report Validation summary Gradebook synthesis Source repo

System Story

From reporting need to reviewable decision

01

Problem

Assessment teams need growth, completion, skill, and data-quality views, while a public portfolio cannot expose real students, rosters, grades, LMS exports, or institutional reports.

02

Approach

Python and DuckDB build validated synthetic warehouse records; SQL produces shared extracts; R and browser code consume those extracts for reports and interactive analysis.

03

Result

The same published data supports course, grade, teacher, section, growth, completion, skill, distribution, readiness, and reporting views without duplicating analytic truth.

04

Lesson

The privacy boundary works best when generation, validation, SQL export, dashboard behavior, and reporting language are designed as one release pipeline.

Proof Lanes

Three proof lanes

Stakeholder dashboard

The dashboard runs on GitHub Pages from SQL-shaped synthetic extracts, with browser-side aggregation, SVG charts, filters, and decision notes tied to visible metrics.

Open dashboard

Decision guardrails

Performance and participation remain distinct, ranked insights exclude cohorts below the default minimum sample size, and validation checks the public data boundary before release.

Read R report

Reproducible analytics

SQL, Python, and R workflows produce course-section performance, growth, non-participation, roster reconciliation, readiness, modeling, and validation artifacts from the synthetic DuckDB warehouse.

Simulation foundation

Safety Boundary

Synthetic by design

Public artifacts must not include real student names, emails, IDs, grades, rosters, LMS exports, submissions, or private school reporting artifacts. Canvas is treated as one possible adapter, not the project identity.