Ambees

Ambees is an adaptive learning platform for students preparing for UTBK, Indonesia's university entrance exam.

It connects a study roadmap with practice, explanations, and feedback on what to work on next.

Role
full-stack
Timeline
July 2026 - present

Finding a direction for exam preparation

Preparing for UTBK means deciding what to study as well as answering questions. Ambees supports students who want a clear roadmap rather than choosing practice topics at random.

From a roadmap to practice

The roadmap moves from fundamentals toward tryout simulations. Students can use the question bank, adaptive tryouts, and explanations alongside learning materials. These tools connect planning, practice, and review within the same product.

Turning mistakes into the next session

Ambees uses answer patterns, frequently missed topics, time spent, and question difficulty to recommend more focused practice. Its progress views show topic performance and question-solving speed. The charts on the public website illustrate these views; they are not measured outcomes from this case study.

My contribution

My full-stack work covered three parts of the platform 1:

  • I implemented a deliberate practice engine using interleaving, spaced repetition, and rule-based question generation.
  • I designed and built a multimodal chatbot with streaming text responses and background processing for intensive AI tasks.
  • I handled in-app transactions for consumable credits through Midtrans.

These contributions cover practice scheduling, AI interaction, and the credit transactions that support paid usage.

Access and credits

Students can begin with free practice. Paid access uses one-time QRIS payments without automatic renewal. Emerald credits provide access to additional AI video generation and tryouts.

Sources

Product descriptions are based on the Ambees website and feature overview. The contribution details come from my existing portfolio record.

Footnotes

  1. This final example uses TfidfVectorizer's built-in tokenizer and English stop-word filtering. The NLTK tokenizer shown earlier is a separate example and is not called by this version. See the TfidfVectorizer documentation. ↩