Systems portfolio

Grant McCurdy

I build data systems, analytics products, AI-assisted workflows, and source-grounded tools that turn messy information into usable decisions, artifacts, and operations.

Explore Projects Open Demos

Start with the dashboard, then inspect the source-backed project briefs and public-safe evidence packets.

Data Systems Structured data, validation, reporting marts

Analytics Products Dashboards, extracts, decision support

AI Workflows Human review, feedback, automation gates

Content Intelligence Corpus, OCR, transcripts, source-grounded reports

Start Here

A simple map of the public work

Use the project directory for the full portfolio, open the demos for working artifacts, or read the case studies for the system story behind selected builds. Source links live with each project so the homepage stays navigable.

Project Portfolio

Six public systems to inspect

Systems data foundation Education Data Simulation Engine

Public-safe simulation, validation, SQL-ready data, and reporting marts for analytics prototypes.

Analytics and decision support Assessment Intelligence

SQL/R/Python extracts, dashboards, diagnostics, and reporting patterns that translate records into reviewable decisions.

Statistical modeling Growth Analytics in R

Adjusted BOY/EOY growth modeling, section signal diagnostics, validation, and stakeholder-facing decision support.

Workflow automation Assessment-to-Remediation Pipeline

Diagnostic design, review gates, dry-run payloads, feedback, follow-up actions, and evidence tracking.

Content systems Content Intelligence

Source-grounded corpus, transcript enrichment, method packs, and report generation from messy source material.

Human-reviewed AI workflows Instructional AI Workflows

Rubric evidence, review packets, feedback drafts, and action artifacts with human judgment kept in the loop.

Open Project Directory

Statistical Methods

Statistical methods behind the analytics work

This supporting packet documents the methods I use across the portfolio: nonlinear model search, Fourier terms, GLMs, repeated cross-validation, calibration, thresholds, and residual diagnostics. The artifacts are public-safe derivatives of graduate statistics coursework.

Open Methods Evidence Skills Index Source repo

01 Model Search

Polynomial, sinusoidal, Fourier, and kernel smoothing benchmarks with complexity tuning.

02 Validation

Repeated stratified cross-validation, log-loss selection, Brier score, AUC, AIC, and BIC.

03 Diagnostics

Residual behavior, Q-Q checks, leverage, spread, calibration, and extrapolation plausibility.

04 Decision Metrics

Sensitivity, specificity, PPV, NPV, threshold analysis, and interpretation-focused sensitivity models.

Demos

Open these first

Interactive dashboard Assessment Analytics Dashboard

SQL-extract-backed synthetic assessment data, browser-side aggregation, SVG charts, filters, and decision notes.

Analytic workspace Portfolio Data Lab

Structured chat surface for tables, summaries, charts, and cited analysis over a public-safe warehouse.

Demo report Content Intelligence Demo Report

A source-grounded reporting pattern built from synthetic notes and visible evidence labels.

Case Studies

Public-safe examples with visible evidence

01 Analytics Reporting Architecture 02 Feedback, Review, and Action Artifacts 03 Content Intelligence Pipeline

Operating Principles

Practical systems, public-safe evidence

Public-safe by default

Demos use synthetic, generated, or permission-safe data so the work can be inspected without exposing private records.

Human-reviewed automation

AI and scripts help draft, structure, and accelerate work, while review gates keep judgment and accountability explicit.

Source-grounded reporting

Outputs should trace back to declared source material, reproducible processing, and clear privacy boundaries.