Owning the AI roadmap for Pandai Teacher. Leading 3 engineering squads alongside design, QA and data. Reporting to Head of Product and CTO.

• Launched a RAG question generator grounded in a 1M+ curriculum-tagged bank, now the source of 53% of all assessment quizzes; owned distractor-quality sign-off across a 57-case acceptance matrix.
• Shipped an agentic teacher assistant that chains curriculum mapping, lesson planning and web search from one prompt; reached 29% weekly active adoption in two months, with confirmation gates and accuracy-over-quantity rules that made agent output safe to trust.
• Built an AI OCR ingestion pipeline turning scanned and handwritten exam papers into digital question sets with LaTeX maths and auto-generated answer schemes; 45% of active teachers now use it.
• Drove 12% free-to-premium conversion by designing an AI-quota paywall on top of a weighted at-risk student model built from dormant quiz-submission data; hit 47% weekly active use among paying teachers within three months.
• Rebuilt the AI generation suite around 10 new generators after deprecating overlapping tools; lifted weekly recurring usage to 26%, exceeding the 20% target, with 41% of AI-generated assessments exporting directly into Quiz Builder.
• Specced the AI Grader MVP for handwritten Paper 1 (non-essay) scripts across 6 subjects, with partial-credit logic and teacher review of low-confidence marks; pilot hit 93% agreement with human markers, scoped to cut grading from 10+ hours to under 30 minutes per class.
• Secured RM 1,000,000 in national grant funding by contributing the AI teacher-assist product design to the Program Anak Kita (PAK) submission.

Owned product delivery for teacher-facing AI tools and core platform UX, working across engineering, design and data. Reported to Head of Product.

• Built and launched 13 generative AI tools for educators on an internal API builder; ran prompt refinement and API workflow optimisation with engineering to hold output quality consistent across every feature.
• Cut tutor support enquiries 30% and lifted tutor satisfaction 25% by shipping the Pandai Tutor Dashboard covering attendance tracking, payment history and leave management.
• Grew monthly active users 75% to 35,000 in 4.5 months by leading a UX revamp of the learning platform grounded in behavioural analysis and data-driven design decisions.
• Shaped the classroom lesson feature from user research into a prioritised backlog; now used across 1,600+ classes, with 41,000+ students completing teacher-assigned lessons.

• Improved tuition operations efficiency 60% through user interviews and process changes born from front-line work.
• Coordinated 130+ live classes weekly and supported 500+ students daily across Telegram and WhatsApp.