Assessment Intelligence

Stakeholder-facing assessment analytics built from public-safe synthetic data.

Project Links

• Open the live assessment dashboard
• Read the concise portfolio brief
• Review the reporting case study
• Inspect the rendered R synthesis report

Assessment Intelligence turns SQL-backed assessment records into an interactive dashboard, reproducible reports, data-quality checks, and decision notes that help instructional leaders answer four practical questions:

• Are students improving over time?
• Where are performance or participation signals weakest?
• Which courses or sections warrant closer review?
• Is the underlying data reliable enough to support a decision?

All public artifacts use synthetic students, teachers, sections, assessments, and outcomes. No real student, school, LMS, personnel, or employer data belongs in this repository.

The first dashboard view presents current score, growth, completion, proficiency, distribution, and readiness signals. Ranked decision insights use a minimum cohort size of 10 so small groups are not elevated as headline findings.

What The Project Proves

Dashboard and stakeholder interpretation

The deployed static dashboard provides browser-side filtering and aggregation, SVG charts, section detail, and generated decision notes over a public-safe SQL extract. The dashboard presentation is deployed from the portfolio repository:

• Dashboard HTML source
• Dashboard JavaScript source
• Dashboard logic smoke test

This repository owns the assessment extract contract, dashboard-data builder, statistical workflows, validation, and reporting artifacts. The portfolio repo owns the shared visual shell used to publish them.

The builder also writes a publication manifest containing the dashboard SHA-256, builder SHA-256, extract hashes and row counts, and dashboard record counts. The portfolio site's quality check verifies its deployed JSON against that manifest, making source drift visible before publication.

SQL-backed analytics

Five SQL extracts support distinct stakeholder questions:

Extract	Decision use
Course-section performance	Compare score, proficiency, completion, and participation signals.
Assignment growth	Identify course-level movement using matched observations.
Non-participation	Keep missing participation separate from achievement evidence.
LMS reconciliation	Confirm roster records are reportable before analysis.
Student readiness	Support curated readiness and observed-growth views.

DuckDB is the reproducible local baseline. Supabase is an optional hosted serving layer over selected public read-only views; it is not required to review or rebuild the local demo.

Reproducible statistical reporting

The R workflows generate synthetic assessment records, fit growth and completion models, reconstruct a public-safe Canvas-style gradebook, validate distribution fidelity, and render stakeholder and analyst reports. The Python workflows extract SQL marts, build dashboard JSON, validate privacy boundaries, and generate deterministic report artifacts.

Architecture

education-data-simulation-engine
  synthetic Canvas-style records and DuckDB marts
                |
                v
assessment-intelligence
  SQL extracts -> validation -> analysis -> reports -> dashboard JSON
                |
                v
grant-mccurdy.github.io
  static dashboard presentation and portfolio case study

The separation is intentional:

• education-data-simulation-engine owns public-safe source simulation.
• assessment-intelligence owns assessment analytics and interpretation.
• grant-mccurdy.github.io owns the deployed presentation shell.

Reproduce The SQL Reporting Path

Build the sibling synthetic warehouse:

cd ../education-data-simulation-engine
make analytics-install
make warehouse

Generate the assessment extracts and analyst report:

cd ../assessment-intelligence
make analytics-install
make sql-warehouse-report

Build the dashboard JSON and publication manifest, then sync both portfolio deployment artifacts:

make dashboard-sync

Run the public dashboard privacy validator:

make all

The validator checks required fields, forbidden identity fields and values, synthetic ID conventions, percentage bounds, cohort sizes, and disclosure of the private-to-synthetic boundary.

Secondary Evidence

The repository also contains deeper proof for technical reviewers:

• sql/extracts/ and sql/extracts_postgres/ define the DuckDB and hosted extract contracts.
• scripts/build_sql_dashboard_json.py converts those extracts into the dashboard data model and hash-bound publication manifest.
• data/published/assessment-dashboard.manifest.json records the exact SQL extracts, builder, counts, and dashboard payload used for publication.
• reports/sql_warehouse_assessment_report.md provides a generated analyst brief with low-sample ranking safeguards.
• reports/gradebook_reconstruction_validation.md documents synthetic gradebook fidelity checks.
• outputs/plots/ contains decision-oriented raincloud, hexbin, and calibration examples.
• docs/privacy-model.md and docs/synthetic-data-methodology.md document the public/private boundary and generation method.

Large reproducible local outputs, including the full dashboard JSON and long-form score table, are intentionally ignored here. Reviewer-facing copies are either summarized in committed reports or deployed through the portfolio site.

Repository Layout

assessment-intelligence/
|-- analysis/          # R generation, modeling, reconstruction, and validation
|-- R/plots/           # Decision-oriented statistical visualizations
|-- data/synthetic/    # Reviewed compact synthetic evidence
|-- data/external/     # SQL-backed public-safe extracts
|-- docs/              # Methodology, privacy, and data contracts
|-- outputs/plots/     # Rendered public-safe plot examples
|-- reports/           # Analyst, validation, and R Markdown artifacts
|-- scripts/           # Extraction, dashboard export, privacy, and report tools
|-- sql/               # DuckDB and PostgreSQL extract queries
`-- screenshots/       # Recruiter-facing dashboard evidence

Public-Safety Boundary

Public artifacts must not contain real names, emails, IDs, grades, rosters, submissions, school records, private LMS links, credentials, tokens, or raw institutional exports. Private-reference workflows may inform synthetic shape and validation rules, but private rows and identifiers remain outside this repository.

Current Status

The public dashboard, shared SQL extract path, publication manifest, R reporting workflows, optional hosted extract path, privacy validator, and reviewer-facing evidence are implemented. The portfolio site consumes the generated dashboard and rejects payloads that do not match the publication manifest.

Licensing

• Code is available under the MIT License.
• Original documentation and generated visual content are available under CC BY 4.0.
• Original synthetic datasets are available under CC BY 4.0.

Third-party materials, trademarks, personal likenesses, and acquired source material are excluded unless explicitly stated otherwise.