Case study · Education
A bilingual AI training program, deployed across a multi-school system.
01 — Challenge
Teachers had an AI mandate, no training, and classrooms where connectivity can't be assumed. Off-the-shelf courses assumed English, bandwidth, and free time — all three wrong for this system.
02–03 — Analysis → Solution
The constraint became the architecture: a complete six-week program engineered as a single offline-capable HTML file — bilingual with full right-to-left support, self-paced, zero installs. If it opens in a browser, it runs.
04 — Architecture
One file. Whole program.
- MODULE ENGINE
Six weeks, gated
Sequenced modules with teacher and administrator tracks splitting at weeks 3–4.
- LANGUAGE LAYER
EN ⇄ UR, full RTL
One codebase, two languages at parity — layout mirrors correctly, not approximately.
- OFFLINE CORE
No server needed
The entire program travels on a USB stick if it has to. Schools run what they have.
05–06 — IMPLEMENTATION & TECHNOLOGY: single-file HTML/CSS/JS · ~471 KB · Apex Prism design tokens
07–08 — Results & ROI
What shipped, counted honestly.
Multi-school
live rollout, government-supported
2 languages
one codebase, full parity
0 installs
runs on what schools already have
ROI framing: delivered as self-paced software, the per-teacher cost is a fraction of conference-style PD — with no substitute-teacher days lost. Exact figures published when the rollout completes its first term.
09 — Lessons learned
The part most case studies delete.
Our post-launch audit found real gaps: progress didn't persist between sessions, there was no submission channel for deliverables, and certificates could be reached without earning them. We're fixing all three — and publishing this because a case study that admits nothing teaches nothing. Clients who read this far deserve the truth about how shipped software actually matures.