Where I prototype the future of AI-assisted design.

Not client work, experiments. AI as a design material, not just a feature to design for.

Lab 01 - AI Strategy · Luxury E-commerce · Dubai · 2025

Aura

AI-accelerated personalisation and pre-loved integration for Gen Z luxury shoppers in the Middle East.

Aura is a strategic design concept reimagining luxury e-commerce for the Middle East. Through discovery research including behavioural data analysis, market studies, and UX benchmarking, I identified a strategic gap: Gen Z luxury shoppers have the highest conversion rates on platform, but pre-loved luxury, their highest value priority, was invisible on the homepage.

What would take weeks of manual persona research was accelerated, modelled, and validated in hours with AI deep research tools, producing "Layla," a data-driven composite persona rather than a fictional character.

AI-powered modules designed

Curate My Closet Remixes past purchases with new and pre-loved pieces to complete a look.

Style My Closet Acts as a personal style coach, analysing uploaded wardrobe items.

Style Stack After adding to cart, suggests curated pre-loved accessories under a set budget.

Lab 02 - Healthcare AI · Prompt Engineering · LLM Architecture

The Clinical Glance

How a language model transforms fragmented patient records into a glanceable clinical card.

A physician has 60-90 seconds to scan a patient's history before entering the consultation room. The source data is raw EHR: unformatted, inconsistent, and dense.

I designed a prompt engineering architecture, The Clinical Glance, that uses an LLM as a Clinical Logic Engine. The system extracts only what's clinically relevant, surfaces critical flags (allergies, missing data, unresolved conditions), and outputs a clean, structured Patient Summary Card. No hallucination. No interpretation. Clinical judgment stays with the physician.

Design principles

Zero filler No introductory text, no hedging, output begins immediately with the first JSON key.

Alert first Overdose risks, unresolved allergies, and missing critical data surface before any summary.

Confidence calibration The model does not diagnose or interpret. It surfaces, and flags what's missing.

Human layer One patient-stated concern extracted verbatim. Clinical judgment stays with the physician.

Status: Concept prototype · Ongoing