# Musaed

> An Arabic and English document assistant helping UAE families of Students of Determination understand information, organize plans and prepare cited drafts locally.

Website: https://zencode.my/work/musaed/
Industry: Inclusive education & family support
Project context: Local bilingual AI application
Services: Local AI, Bilingual product design, Document intelligence, Guided workflows

## Complex documents. A clearer next step.

### The starting point

Families may need to work through reports, school information, insurance documents and application requirements in more than one language. They need help understanding what a document actually says, and a manageable way to prepare their next step.

### The approach

Musaed combines a calm bilingual interface with a local document and retrieval pipeline. Answers connect back to source excerpts. Guided journeys organize tasks and linked documents, while draft emails and meeting preparation stay under the family’s control.

## What went into the build

### A human starting point

Understand a report, get help with school or insurance, or start a conversation—without having to learn the technology first.

### Arabic and English, thoughtfully handled

Language-aware navigation, text direction and bilingual drafts support a mixed-language document experience.

### Answers with something behind them

An evidence viewer shows source excerpts and page references. Unsupported requests can be declined instead of answered speculatively.

### Documents stay close

Parsing, retrieval and generation run locally, with household documents and saved plans stored on the device.

### From reading to preparing

Family journeys organize tasks, linked reports, confirmed dates and preparation for school or insurance conversations.

### Control over the next step

Drafts and journey packs can be reviewed and downloaded. Messages and applications are not automatically submitted.

## Scope

- Document reading
- Evidence & citations
- Arabic / English
- Family journeys
- Draft preparation
- Local storage


## Workflow

1. **Bring a document:** Upload a report or source document and check that its text can be read.
2. **Ask a question:** Search the local document collection for relevant supporting passages.
3. **Check the evidence:** Read the answer alongside its retained quotations and source references.
4. **Plan a next step:** Organize a journey or prepare a draft for the family to review.


## Engineering decisions

### Privacy shapes the product

The core document and AI workflow runs locally. That choice is reflected in storage, model use and the interaction design.

### Evidence is part of the interface

Citations are inspectable. Missing evidence is a reason to ask for more information, not to manufacture a confident answer.

### Assistance without taking over

The application helps families prepare. It does not make clinical, eligibility or insurance coverage decisions on their behalf.


## Built with

Python, Streamlit, Local language models, Multilingual retrieval, Chroma, SQLite

## Project context

The interface previews contain illustrative text, not family reports. Musaed is an assistive document tool; it does not make clinical, eligibility or coverage decisions.

Project enquiry: https://zencode.my/contact/