Back to Homepage

Highlights

Students & Internships

Over a ten-week industrial attachment at the AI Unit, hosted at the Mauritius Emerging Technologies Council in Ebene, a team of university interns redesigned and rebuilt aim.govmu.org, the official government portal for artificial intelligence in Mauritius. The work spanned front-end development, WCAG accessibility, speech synthesis and the integration of the DIVA chatbot. This page is a record of what was built.

The internship cohort of eight students standing together in the AI Unit office.
The intern team at the AI Unit, METC, during the May to July 2026 attachment
Scroll to explore Go to the internship overview
An intern presenting a diagram titled The Big Picture to colleagues seated around a meeting-room table.
Walking the team through the big picture: the front end and the planned FastAPI and PostgreSQL back end

The internship

A live national portal, built by interns

The AI Unit had no permanent development staff, so its active projects were carried forward by the intern team. The placement began with a single intern analysing the existing portal and setting up the project; by the later weeks it had grown into a team of several interns from different Universities working in one shared codebase.

  • Collaboration

    A flat, collaborative structure with the mentor as the single source of technical guidance, and work coordinated through a shared Kanban board.

  • Teamwork

    New interns were onboarded onto the existing code and workflow, features were divided across the team, and pieces like the PDF-to-audio converter were built and integrated by different interns working together.

  • Learning by doing

    WCAG standards, the Web Speech API and Jira were all learned through self-directed research and then applied directly to the portal, with the mentor guiding rather than handing over solutions.

  • Government digital projects

    The portal serves a national audience, and the work was presented to the Ministry of Technology and to the Electoral Commission of Mauritius during the attachment.

Project one

AI Unit Website Redesign

aim.govmu.org is the official Mauritian government portal for artificial intelligence, covering news, policies, events and resources.

The rebuilt Framework Library page, showing cards for the Digital Transformation Blueprint and Mauritius' First National AI Strategy, each with download, view online and listen options.
The rebuilt Framework Library. The Listen buttons are the PDF-to-audio feature: the AI Strategy, FAIR Guidelines, Digital Blueprint and AI Playbook can be read aloud.

From analysis to wireframe to build

The work began with a full review of the existing portal, guided by the AI Unit's aspirations for its redevelopment. That analysis fed a detailed wireframe covering the home page, the navigation structure, the key content sections and the placement of interactive features such as the accessibility tool, and the chatbot. Once the wireframe was reviewed and approved by the mentor, front-end development began.

  1. 01

    Analysing the existing site

    Review of aim.govmu.org: styling, mobile-friendliness, accessibility gaps and missing features.

  2. 02

    Wireframing

    A detailed wireframe of the home page, navigation and feature placement, approved before any code.

  3. 03

    Modern UI/UX

    A consistent design language built on typography, colour schemes and spacing.

  4. 04

    Responsive development

    A fully responsive layout built with CSS Flexbox and Grid.

Frontend

Built with HTML5, CSS3 and JavaScript (ES6+) without any front-end framework, so the codebase stays easy to maintain for future developers.

Back-end planning

Python with FastAPI and PostgreSQL was agreed as the back-end stack, with API design patterns and database schema research under way at the end of the attachment.

Responsive design

A fully responsive layout using CSS Flexbox and Grid, with a consistent scheme of typography, colour and spacing.

A physical Kanban whiteboard with columns for Backlog, In Progress, Test, Blocked and Done, filled with printed task cards.
The physical Kanban board that came first, before the workflow moved to Jira

Kanban, Jira and a shared repository

Project management for the portal team was run on a Kanban workflow, introduced in Week 2 after guidance from the mentor on Agile practice. It began as a physical board at the workplace and, at the mentor's recommendation, moved to a digital Jira board that the whole intern team adopted as it grew. Tasks were grouped into Epics and Sub-Epics matching the portal's features.

GitHub served as the team's version control platform, with a shared repository that let all the interns work concurrently on different features without conflicts in the code.

  • Backlog
  • In progress
  • Test
  • Blocked
  • Done
  • MethodAgile, Kanban workflow
  • TrackingJira board, Epics and Sub-Epics
  • Version controlGit and GitHub, shared repository

Project two

DIVA - Digital Interactive Virtual Assistant

DIVA is an AI-powered assistant integrated into the portal to answer questions about artificial intelligence in Mauritius, the content of the portal and government AI resources. It is a prototype, grounded in three key documents: the Digital Transformation Blueprint, the AI Strategy and the FAIR Guidelines.

The Meet DIVA introduction card, describing the assistant as answering questions based on four key documents, with a Chat with DIVA button.
The invitation to start a conversation, placed alongside the documents themselves

Two ways into the conversation

Users reach DIVA through two entry points. A dedicated section of the portal introduces the assistant and opens it with a Chat with DIVA button, and a floating chatbot widget stays available across the portal, so a conversation can be started or continued from any page without leaving the current content.

  • AI assistant

    Answers questions on AI in Mauritius, the portal's content and government AI resources, drawing on the four published documents.

  • Integration work

    The first integration attempt failed against the portal's Content Security Policy. The conflict was diagnosed through browser developer tools and resolved with a CSP configuration that permits the required cross-origin calls while keeping the portal secure.

The DIVA chat window showing a detailed answer about the Digital Transformation Blueprint, with read-aloud and copy controls and a question box offering typed or spoken input.
The conversation view, with read-aloud, copy and voice input

Speech synthesis: a portal that reads aloud

The Web Speech Synthesis API, a native browser API that turns text into speech without any third-party software, was integrated into the portal. Activated with a single control, it reads the content of DIVA's response aloud.

  • Conversational interface

    Questions can be typed or spoken, answers can be read aloud or copied, and the conversation can be cleared and restarted at any point.

  • Accessibility

    Reading page content aloud supports users with visual impairments and lower literacy levels, in keeping with the WCAG principles behind the portal.

Project three

Accessibility

In Week 3 the mentor raised the bar: the portal had to meet WCAG 2.2, the international guidelines for making web content usable by people with visual, hearing, physical, cognitive and other disabilities. Beyond meeting the guidelines, the interns built an accessibility toolbar into the portal itself, designed on the principle of user choice: each visitor decides how they want to use the site.

The accessibility panel's screen reader section, offering read page aloud with play, pause and speed keyboard shortcuts, a voice selector, and quick profiles for low vision, motor, dyslexia, cognitive and senior needs.

Screen reader & quick profiles

A built-in screen reader reads the page in natural segments while highlighting the current element and scrolling it into view. Space pauses, S stops, arrow keys change the speed, and a slider and voice list give finer control. Five quick profiles, for low vision, motor, dyslexia, cognitive and senior needs, apply a set of adjustments in one click.

The accessibility panel's text size control and colour and display options, including normal, high contrast, dark, greyscale and negative modes, with toggles for highlight links, hide images and stop animations.

Text size & colour

Base text size scales between 80% and 150% without breaking the page layout. Five display modes, Normal, High Contrast, Dark, Greyscale and Negative, cover different contrast needs, alongside toggles to highlight links, hide images and stop animations.

The accessibility panel's reading and focus options - dyslexia-friendly font, reading guide line, wider letter spacing and bold focus outline - plus navigation aids and the Alt plus A keyboard shortcut to open the panel.

Reading, focus & navigation

A dyslexia-friendly font toggle, a reading guide that follows the cursor, wider letter spacing and a bold focus outline for keyboard navigation, plus a large mouse pointer for users with motor difficulties and a keyboard shortcut guide. The panel itself opens with Alt + A.

The standards behind the switches

Meeting WCAG 2.2 meant independent research into the A, AA and AAA conformance levels, testing the portal with browser-based accessibility evaluation tools, and reworking front-end code that had already been written. It was a practical lesson in building accessibility in from the start rather than retrofitting it.

  • WCAG 2.2

    The portal was tested against the guidelines with browser-based evaluation tools, and the required adjustments were identified and applied across the front-end code.

  • Colour contrast

    Contrast ratios in the existing front end were reviewed and improved to meet the guidelines, with a High Contrast display mode available for users who need more.

  • Keyboard navigation

    Keyboard navigation support across the portal, a bold focus outline to keep the current position visible, and a shortcut guide for fully mouseless use.

  • Screen readers

    The toolbar announces its state changes through an ARIA live region, keeping external screen readers such as NVDA and JAWS in step with what is on screen.

  • ARIA labels

    ARIA labels and text alternatives for non-text content were added throughout the front end so that assistive technology can name and describe every control.

  • Saved preferences

    Every setting is stored in the browser's local storage, so a returning visitor's preferences are restored automatically with no account needed.

Portrait of Mrs Aarthi Burtony André, SENA Resource Person for Learners with Visual Impairments.

Mrs Aarthi Burtony André

SENA Resource Person for Learners with Visual Impairments

Date
Venue
SENA Office
Tools
NVDA, Microsoft Narrator

Accessibility testing with SENA

Mrs Aarthi Burtony André, SENA Resource Person for Learners with Visual Impairments, took part in accessibility testing of the website at the SENA Office on 24 July 2026, at the request of two AI Unit interns. The website was tested with the NVDA and Microsoft Narrator screen readers while it was still under development.

Her feedback and recommendations from the session were taken into consideration, and changes were subsequently made to improve the accessibility and usability of the website.

Mrs Burtony André welcomed the Ministry's initiative in engaging with persons with disabilities during the development of its AI Unit website, and described the collaborative approach as greatly appreciated and as reflecting good practice in inclusive digital design. The AI Unit thanks her and SENA for taking part in this testing, which continues to inform accessibility work on the website.

The process

Development journey

The same six stages carried every piece of work from an open question to something ready to hand over.

The Child Safety Series video player showing an animated film with French captions, an English and French language switch, and a captions toggle with a language selector.
The video captioning feature: audio transcribed by hand into timed WebVTT files, surfaced through the HTML5 track element, with captions available in English and French. The interface presenting the videos and captions was also built by the interns.

  1. Stage 01

    Research

    Analysis of the existing aim.govmu.org site, plus self-directed research into WCAG 2.2, the Web Speech API and open data.

  2. Stage 02

    Planning

    A wireframe of the home page, navigation and feature placement, approved by the mentor, alongside the Kanban board that would organise the work.

  3. Stage 03

    Development

    Front-end build in HTML5, CSS3 and JavaScript, followed by the accessibility toolbar, speech synthesis, DIVA, video captions and PDF-to-audio.

  4. Stage 04

    Testing

    The portal was checked against WCAG 2.2 with browser-based evaluation tools, and integration problems, like the chatbot's CSP conflict, were diagnosed through browser developer tools.

  5. Stage 05

    Refinement

    Early front-end code was reworked to meet the WCAG requirement, and the speech engine gained a voice selection algorithm to align voices with the chosen language.

  6. Stage 06

    Deployment preparation

    The portal was presented to the Ministry of Technology, back-end planning began, and the code structure was documented so incoming interns could continue the work.

Interns and a member of the unit standing in the Ministry of Information Technology, Communication and Innovation reception, beside a plaque reading Transform Mauritius into a thriving, smart and inclusive digital society.
At the Ministry of Information Technology, Communication and Innovation, where the portal was presented.

Gallery

AI Unit Internship

Moments from the AI Unit internship, including website revamp, accessibility, DIVA and presentation work.

Rodrigues & Imperial College Programme

Programme highlights and activity images for Rodrigues and Imperial College collaboration work.

Capability

Skills & technologies

The technologies used across the portal, and the tools the team worked with.

  • HTML5

    Semantic markup, ARIA labels and the track element for captions

  • CSS3

    Responsive layouts with Flexbox and Grid, and a consistent design language

  • JavaScript (ES6+)

    All portal interactivity, written without a front-end framework

  • Frontend Development

    Wireframing, interface build and usability improvements

  • Responsive Design

    A layout that holds from a small phone to a wide desktop

  • Accessibility

    WCAG 2.2, ARIA labels, keyboard navigation and contrast

  • Web Speech API

    Speech synthesis for the screen reader and PDF-to-audio

  • AI Chatbot

    Integration of the DIVA assistant into the portal

  • Python

    Selected for the back end, researched during the attachment

  • FastAPI

    The agreed back-end framework, with API design patterns explored

  • PostgreSQL

    The chosen relational database, with schema planning under way

  • Git & GitHub

    A shared repository for concurrent work across the intern team

  • Jira

    Kanban board with tasks organised into Epics and Sub-Epics