AvaB

Good teaching does not scale on its own. A great teacher can reach one room; the same lesson, personalized to hundreds of different learners, is a technology problem. AvaB began with one question: how can technology make learning more personalized, engaging, and actionable — without lowering quality?

I had spent years as a competition student, optimizing for the right answer. Building AvaB forced a harder question: does the learner on the other side of the screen actually understand, or did we just make a prettier worksheet? I wanted to turn what I knew about mathematics and problem-solving into something a real student uses on an ordinary Tuesday night.

As team leader I set product direction and coordinated four teammates. Concretely: I shaped the learning flow, worked on the personalization logic, engineered content (a grade-1 exercise set on the platform credits me directly as its author — 'Người soạn: Đỗ Bảo Nam'), and ran early user testing. I am documenting exactly what I built versus what the team built so this page states it precisely.

AvaB pairs a content library with a personalization layer: learners are routed to the next item based on what they got wrong and why, not just whether they were right. AI assists with generating variants, explanations and hints, while human review keeps the pedagogy sound. The platform runs live for real teachers and classes.

avab.vn is a live platform with real learners, teachers and paid courses: around 1,500 learners and parents, roughly 230 paid enrolments a year, and about 350M VND in annual revenue. As a student project it won first place at Newton, second across the Nam–Bắc Từ Liêm cluster and third in Hanoi at the student-startup competition, and was selected for the National Student Startup round. These figures come straight from avab.vn's own user and revenue records, and the competition results are on the record.

The hardest calls were not technical. Deciding what to cut for a first release, splitting work across teammates with different strengths, and holding a quality bar when it would have been faster to ship something shallow — those defined the project more than any feature did.

Our first personalization rules were too aggressive: they pushed some learners ahead too fast and lost them. I learned that a model that is 'right on average' can still fail the individual student in front of you — so we slowed the pacing and added human checkpoints.

The platform is public at avab.vn. Authorship of specific content, the SV_STARTUP-2024 record, and analytics are being attached to this page as verifiable evidence rather than claimed in prose.