DSE prep, rebuilt around adaptive practice.
An e-learning platform for HK DSE students: mock papers, study materials, and the studio's first deep integration of Gen-AI into the design loop. Live with DSE students and teachers via the studio's tuition-centre partners. Teacher authoring time compressed from a multi-day PDF-and-Word loop to a same-day in-platform flow, and adaptive practice and self-grading surfaces are used weekly through the exam run-up.
05 · cover plateHanlun · Gen-AI e-learningHK DSE is the public examination that decides where Hong Kong students go to university. High-stakes, repetitive, and historically badly served by static PDFs. The brief: build a platform that serves three audiences inside one product (students practising, teachers authoring, admins overseeing).
Three audiences, three different definitions of 'done.' Students need adaptive practice and clear feedback. Teachers need authoring leverage to scale themselves. Admins need cohort oversight without micro-managing. The challenge was finding the shared spine that didn't dilute any of them.
“We are learning what design judgment means when production speed stops being the bottleneck.”
Two parallel tracks: the product, and the way we built the product.
Student · adaptive practice
Mock papers and platform materials, sequenced by performance rather than by file order. The student surface hides the machinery and just shows the next reasonable thing to attempt.
Teacher · authoring leverage
Teachers can build, edit, and re-issue mock papers without an engineer in the loop. The authoring surface is where most of the AI affordance lives: scaffolding, suggestions, and review patterns that scale a teacher's voice across a class.
Admin · cohort oversight
Admins see the school, not the individual screens. Cohort progress, intervention prompts, and a calm reporting layer so leadership can act on what is actually happening across the platform.
AI in the design loop
The studio's first deep integration of Gen AI into the design workflow itself. We compressed spec to surface from weeks to days, using AI to draft, iterate and stress-test variants the way a junior designer would, while final judgement on every direction stayed with me.
Cantonese-led, bilingual surfaces
DSE is sat in both Chinese and English, sometimes paper-by-paper. The platform reads Cantonese-led for the student and teacher surfaces, with bilingual marking-scheme language so a teacher can author in Chinese and a student can self-grade against the English convention. Tested with teachers across HK tuition centres before the surfaces went live.
“Before this I was rebuilding the same mock paper in Word every term. Now I author it once, the students sit it, and I see who needs me on Monday morning. I stopped working on weekends.”
Adaptive practice · ruled canvas · self-grade against the scheme
Design loop · spec to surface in days, not weeksI started Hanlun assuming the student was the hardest audience to design for. Six weeks in, the content team lead told me the platform would live or die on what the teacher experience felt like at 11 p.m. on a Sunday. He was right. We moved authoring leverage to the centre of the design, and the teacher quote I keep coming back to ("I stopped working on weekends") is what the platform actually shipped for. I would disagree with most "AI changes design" takes I read this year. The AI affordance here didn't compress the design loop. It compressed the production loop. The judgement seat got bigger, not smaller.