Bridge Studio. Design once. Adapt thoughtfully.
A bilingual lesson-design workspace that keeps the educator responsible for the final learning experience.
The specification, wireframes and browser prototype were created for this portfolio. Research and validation below are proposed. The prototype uses authored templates, not a live AI model, translation service or backend.
Problem hypothesis
Educators preparing Hebrew and Arabic activities may spend substantial effort aligning objectives, language level and assessment. Generic AI drafts can look complete while hiding weak pedagogy or translation drift. This is a hypothesis informed by the portfolio context, not a measured customer finding.
Product vision
Help an educator turn one learning intention into a short, reviewable activity in two languages. Make adaptation decisions visible and require a human decision before export.
Who it is for
| Hypothesized user | Need | Core job |
|---|---|---|
| Technology instructor | A focused activity at an appropriate level | Prepare a short session with a clear success check. |
| Bilingual curriculum developer | Terminology and learning-purpose consistency | Compare Hebrew and Arabic, then edit both. |
| Instructional coordinator | Reviewable quality and consistent standards | Check alignment, safety and readiness before use. |
These are provisional role profiles, not research-validated personas.
A narrow, testable first product
MVP scope: one educator, one short activity, one age band, Hebrew and/or Arabic, a formative question and a required review. The prototype supports three authored topics and three grade bands. A production AI version remains a future implementation.
| Priority | Requirement | Acceptance criterion |
|---|---|---|
| Must · brief | Select topic, grade, duration and language. | Missing required input blocks draft creation with an accessible message. |
| Must · structure | Create objective, opening, practice and assessment. | Each draft contains all four sections and a time allocation matching the brief. |
| Must · adaptation | Display Hebrew and Arabic with correct RTL. | Meaning and assessment intent can be reviewed side by side; Latin terms remain readable. |
| Must · editorial control | Edit the draft and complete a review checklist. | Export stays unavailable until objective, language and privacy reviews are checked; edits reset review. |
| Must · export | Download a plain-text activity. | UTF-8 file retains language direction content and the edited draft; includes prototype provenance. |
| Should · traceability | Show draft version and changes. | Future implementation: record edits without collecting learner identities. |
| Later · AI generation | Generate from an approved reference. | Future implementation: model output validated against a schema; unsupported claims marked for review. |
| Out of scope | LMS integration, learner accounts, grading and auto-publication. | No student records, autonomous publishing or unsupported performance claims. |
Key trade-off
A constrained prototype tests the value of structure, language comparison and review without adding model variability. It cannot validate AI quality. An AI pilot would require a separate evaluation set, protected server-side credentials and approved content sources.
User flow & wireframes
Technical and content boundaries
The working prototype runs locally in the browser. Form selections choose authored templates; no prompts, drafts or personal data are sent to a model or server. Edits exist only in the current page session until the visitor downloads them. There is no account, autosave or analytics.
Proposed AI architecture: brief validation → approved source retrieval → structured generation → schema and language checks → educator editing → review → export. Missing source support should produce a visible uncertainty state. Translation must preserve the objective and assessment, not only the wording. Personal learner data should be excluded from prompts.
Validation plan
Proposed, not conducted: interview six educators across instruction, curriculum and coordination, including Hebrew- and Arabic-dominant users. Ask each to bring a non-confidential planning example. Observe the current workflow before showing the concept. Then run a moderated task: prepare a short activity, inspect both languages, improve one section and export it.
Counterbalance the order of manual and prototype tasks, record assistance needed and use independent bilingual reviewers for meaning, age fit and assessment alignment. A small sample is for finding problems, not proving market demand. Stop an AI pilot if source attribution, privacy or language reliability fails the agreed review standard.
| Proposed metric | Operational definition | Initial decision threshold |
|---|---|---|
| Task completion | Participants exporting an acceptable activity without facilitator rescue / attempted tasks. | At least 5 of 6 in the moderated pilot; directional only. |
| Time to reviewable draft | Minutes from completed brief to a draft meeting the quality rubric. | Compare within participant; no promised time saving. |
| Bilingual fidelity | Meaning, objective and assessment scored by two reviewers. | No critical meaning or safety errors in the pilot set. |
| Review effectiveness | Seeded issues found and corrected before export. | All critical issues detected; revise if any are missed. |
| Repeat value | Educators who choose to use it for a second real planning task. | Qualitative follow-up after two weeks; no retention claim yet. |
All thresholds are proposed acceptance criteria, not achieved results. The next investment decision should follow observed value and reliable language quality, not prototype polish alone.