Learning designer focused on evidence-based, accessible learning in the AI era. My projects cover corporate L&D, K-12 and higher-ed curriculum, EdTech, and research skills.
Selected case studies from my instructional design portfolio. Each one sets out the performance problem, measurable objectives, a modality trade-off analysis, a 3D practice scene and an evaluation plan. The clients and scenarios are fictional, and none of the numbers are field results.
| Case study | What the learner does |
|---|---|
| Wind-Turbine Alarm Calls: Restart or Hold? | Offshore wind control-room technicians decide whether to restart or hold across eight alarm cases under production-desk pressure, then write a handover note |
| Cold-Chain Goods-In: Decision Training for Dock Receivers | Chilled-goods dock receivers complete a 21-decision certification module and a dock run against the clock, where core temperatures are hidden and the CMR can be signed too early |
| Phishing Recognition Drill: Report, Verify or Trust | Office staff read each sender address right to left, name the registered domain, and decide whether to report, verify or trust the message |
| Pharmacy Double-Check Training: Errors vs Over-Flags | Hospital pharmacy technicians flag or clear five prescriptions, name the faulty line and the problem, and write the annotation |
| Craft-Dairy Process Training: Expert Interview to Floor-Ready Kit | New dairy operators follow a tank-line routine built from a founder interview, choose the record over the surface cue at four gates, and write the deviation note |
| Vetting Adjudication: Weight the Pattern, Not the Story | Vetting staff tag twelve documents in a dossier, weigh one vivid disclosure against a quiet pattern, and write the conditions or the reason |
| Async Process Handover: Built for the 22:00 Reader | Learners build a handover document on a 15-point budget, then test it in a Friday-night replay that uses real IST/CET clock arithmetic |
| AI-Assisted Customer Email: Verify Before You Send | In an 8-minute 3D micro-lesson, support agents find the three real risks in a confident AI draft, climb a four-rung Verify ladder and make the send decision |
Five browser-based 3D learning games and mini courses. Realistic detail appears only where noticing it is the learning. The real data comes from NASA, USGS and the Smithsonian. There is no tracking and no sign-up, and every project has a text mode and full keyboard play.
Play them all in the Immersive Learning Lab →
| Project | What the learner does | Play |
|---|---|---|
| Lunar Field Expedition | Lands at the real Apollo 15 site, reads the terrain under changing sunlight, plans a sampling traverse and defends the findings | Play |
| The Object Detective | Examines real Smithsonian 3D scans under raking light, checks them against archival records and writes an honest museum label | Play |
| Hazard Hunt | Spots warehouse hazards in realistic 3D, replays a near-miss with pause-and-decide moments and practices reporting (workplace safety) | Play |
| Field Ecologist | Designs a sampling plan for a meadow, then sees the true population and how the sampling design biased the estimate | Play |
| Glacier Time Machine | Aligns USGS repeat photos from 1941, 1950 and 2004 to real 3D terrain of Muir Inlet, Alaska, measures glacier retreat and reasons about causes and uncertainty | Play |
These work best in a desktop or laptop browser, and phones get lighter graphics. Phone support has been checked in emulation only, not on physical phones. Functional, accessibility and performance testing is automated, and agent-based design reviews were run on earlier versions. No real learners have used these yet, and each repo's STATUS.md lists its known open issues.
Each project follows the same arc: analysis → design → build → evaluate. Each one documents its design rationale, its accessibility testing, and an evidence section that states what the evidence can't show. Live sites are on GitHub Pages.
| Project | Problem | Audience | Live |
|---|---|---|---|
| AI Literacy Program | Staff and students need practical, risk-aware AI literacy rather than prompt tips | Employees; students aged 14–16 | Demo |
| Should This Be a Course? | A request for "feedback training" hides environment, process and motivation causes | Corporate L&D, HR | Case study |
| Accessibility Remediation Case Study | A real, openly licensed university course unit audited against WCAG 2.1 AA and remediated, with a plan for scaling to 300 courses | Higher-ed instructional designers, accessibility specialists, faculty | Case study |
| AI-Resilient Assessment Kit | Faculty need assessment redesigns that hold up when students use AI, with explicit AI-use levels | Higher-ed faculty | Kit + workshop |
| Research Tools Academy | Graduate students need task-based Zotero training, built from an analysis of 145 real public support questions | Graduate students | Academy |
| Tutor Training Course | Tutoring quality at scale needs trained tutors, observable tutoring moves and fidelity data | K-12 tutoring programs | Program |
Each tool has negative tests that prove its checks catch seeded errors, and each has a separate agent review (labeled "Agent review") whose findings and fixes are committed in the repo.
| Tool | What it does | The gap it fills |
|---|---|---|
| tracebook | Extracts factual claims from curriculum content, verifies each one against its cited source with quote-back matching, and rewrites approved corrections in place with dated notes. A weekly GitHub Action re-checks every source for link rot | Citations in course content go out of date without anyone noticing |
| slidewalk | Turns one lesson outline into a single-file HTML deck and a native PPTX. Accessibility checks (axe, keyboard, 320 px reflow, contrast, alt text) fail the build instead of warning | Deck generators treat accessibility as a claim rather than a gate |
| chalkreel | An agent skill that turns a measurable learning objective into a short, code-drawn explainer video with captions, a transcript and audio-description notes, checked against written criteria based on Mayer's multimedia principles | AI video generators start from a topic and optimize for engagement, not for a learning objective |
In the GitHub repos, code is MIT and learning content is CC BY 4.0. The one exception is the Accessibility Remediation Case Study, which keeps its source module's license (CC BY-NC-SA 3.0 US). All projects were built with AI assistance as portfolio pieces. Agent reviews are labeled "Agent review", and fictional organizations, personas and simulated datasets are labeled as fictional.