Tupai.ai, an AI-driven math platform by AI Teacher Sdn. Bhd., boosts math skills for Malaysian students through tailored support and progress tracking.
Education website to fosters personalized learning, with metrics tracking its impact on math comprehension for Malaysian students.
0-10 Employees
Tupai.ai is a curriculum-aligned AI math tutor built for KSSM and IGCSE students in Malaysia. The product breaks math concepts into manageable chunks, answers student questions when they get stuck, and provides exam-readiness tooling including past-year papers and question deconstruction.
AI Teacher Sdn. Bhd. partnered with Zensite to design the platform from the ground up: a learning experience that worked for two very different users at once. The student doing the learning, and the parent trying to support that learning without taking it over.
Traditional math education in Malaysia treats students as a homogeneous group. Same pace, same explanation, same examples, same exam preparation. Students who fall behind stay behind. Students who race ahead get bored. And the people most invested in turning that around, the parents, have very little visibility into where their child is actually struggling until a test score arrives too late to do anything about it.
AI Teacher Sdn. Bhd. set out to build a math tutor that adapts to each student individually while keeping parents meaningfully informed. The platform itself had a clear technical path forward. The product experience did not. Designing a single coherent product that serves a 13-year-old wanting to finish a question and a parent wanting to track progress without surveilling their child is a real tension, not a copy-paste UX problem.
The Zensite team led by Kumaran ran the discovery phase across three parallel tracks.
We ran three structured parent interviews, recorded via Fathom, focused on what parents actually want to know about their child’s math learning, what they currently struggle to find out, and how they would react to a dashboard that gave them too much information versus too little. The finding was consistent: parents want clarity, not surveillance. They want to know when to step in, not be told every move their child makes.
Fay attended a teacher workshop hosted by AI Teacher and brought back insights on how Malaysian classroom teachers see AI tools today. Where they trust them, where they don’t, and what gaps in the typical learning experience they would actually want an AI tutor to fill.
We benchmarked against the most direct competitors in the AI math tutor space, including Maths.ai and Answer.ai. Most of the existing field gave students answers. Very few helped students understand. That distinction shaped the product positioning, and eventually the Tupai mission line: We don’t just give answers. We build understanding.
A meta-note on this project worth sharing: we used AI as a research assistant during discovery, and the workflow itself became part of the value we delivered.
Workshop sessions ran 5 to 7 hours each. In the past, turning a session into usable artefacts meant a designer disappearing for two or three days to transcribe, code themes, and write up findings. On Tupai, we recorded sessions in Fathom, fed the structured transcripts through ChatGPT with a tuned prompt library, and used Figma AI to generate initial user-flow diagrams from the synthesis.
The key word here is assistant. AI handled the mechanical work of synthesis. The strategy, the design decisions, the judgement calls about what mattered most in the room, and the editorial taste behind the final flows all stayed with the human team. AI compressed the gap between insight and artefact. It did not replace the people drawing the conclusions.
The product was designed across three primary surfaces.
The student’s home is a chat-based interface with the AI tutor at the centre of the experience. Around it sit motivational elements: streaks, leaderboards, daily goals. Designed to encourage repeat sessions without manipulating young users into addictive loops. A dynamic syllabus tracker shows where the student sits in the KSSM or IGCSE curriculum, what they have covered, and what is next. The design challenge: make the AI feel like a patient tutor, not a search engine. Long answers feel wrong. Step-by-step coaxing feels right.
A supplemental dashboard for students themselves: rankings within their cohort, syllabus progress, and reward claims. Visible enough to motivate, light enough to never feel like a grading interface. Built to be opened occasionally, not stared at.
The hardest design problem of the project. Parents needed enough signal to know when their child was stuck, struggling, or thriving. They did not need a moment-by-moment activity log that turned their child into a tracked subject. We designed the parent dashboard around weekly and monthly trends, key inflection points (concepts the child is repeatedly stuck on, topics where they have unexpectedly improved), and explicit recommendations on when and how to step in. The metric hierarchy was deliberate: parents see learning streaks before they see time spent, and effort before performance.
We built a scalable design system from the ground up: typography, color, components, illustration patterns. It needed to feel warm enough for a child to want to come back to it, credible enough for a parent to trust it, and structured enough to scale to new features as Tupai expanded beyond math.
We delivered an end-to-end product experience for Tupai.ai covering the student learning interface, the student dashboard, the parent dashboard, and the supporting design system. The work shaped the platform that AI Teacher Sdn. Bhd. now operates from Tupai Treehouse in Kuala Lumpur.
In the years since the design shipped, Tupai’s classroom impact has been substantial. Across partner schools, the platform has delivered a 50% reduction in failure rates and an average 1 to 2 grade jump for students using it consistently. Tupai’s own data shows that just 5.5 minutes of daily focused interaction over 3 months is enough to bridge a student from a grade G to a grade C. Two out of every three students using the platform see measurable improvement in their academic performance.
The product has been deployed in schools across Kuala Lumpur, Selangor, Penang, Negeri Sembilan, and Pahang, with expansion to Sabah and Sarawak underway. AI Teacher Sdn. Bhd. has since secured the MDEC Digital Content Grant and completed an equity crowdfunding round on pitchIN, validating both the design and the product strategy with a national funder and a community of investors.
Beyond the artefacts, the deeper outcome was a product positioning that has held up: Tupai is not a "give me the answer" AI. It is a "help me understand" AI. That distinction shows up in everything from the chat micro-copy to the pace of the lesson flows to the way the parent dashboard frames effort over output.
"Zensite is full of truly amazing people. They care deeply about UX design and they go the extra mile to think about how to engage our customers. They revel in putting themselves in other people’s shoes and have met every timeline. Go drop them a message if you’re in need of UX help."
Edmond Yap
Founder, AI Teacher Sdn. Bhd. (Tupai.ai)
Reduction in failure rates across partner schools
Grade jump average for students using Tupai consistently
Students seeing measurable academic improvement
We’d love to learn more about your company and how we can help you. Tell us about your project in the form, and we’ll put you in touch with the right team to get an estimate or quote.
If you have a RFQ or brief, please send to sales@zensite.co
If you are looking for a retainer based service, please check out our design subscription.
.webp)
