My role
Full-Stack Developer · end-to-end
Project year
2026
Original scope
2 months
Context
Personal product
NaraiSensei — public website screenshot
Original screenshot of the public website, captured 15 November 2026. The live product may change over time. Visit the live site

Overview

NaraiSensei brings Japanese language learning and career preparation into one platform for Indonesian learners. The original project brought together a JLPT N5–N1 curriculum, AI-assisted reading and speaking practice, and a career pathway connected to training institutions (LPK).

I designed and built the platform end-to-end on WordPress: information architecture, custom content types, learning flows, user accounts, and AI integration. The aim was to replace scattered PDFs and videos with a guided path that makes the next step clear.

The application also includes scenario-based AI conversations, speaking and interview evaluation, progress reporting, and a WooCommerce-linked access model. These features sit within the learning flow rather than a standalone chat interface.

The challenge

Learning materials were fragmented and progression was unclear. Learners had limited feedback on reading and speaking, while finding reliable job information and training partners added another disconnected process. The platform needed to organize five JLPT levels, support maintainable publishing, and connect study with career preparation.

My contribution & approach

Custom content types organize the curriculum by level, unit, and skill. This gives authors a consistent structure and helps learners understand what to study next.

AI-assisted reading evaluation and speaking practice provide feedback within the learning flow. Career information and an application flow connect learners with LPK partners. A mobile-first interface supports studying on a phone.

Architecture & technical decisions

A speech-to-feedback pipeline

Shadowing practice sends recorded audio for speech-to-text transcription, then evaluates the transcript against the reference using a language model with structured JSON output. The result includes scores and Japanese/Indonesian feedback.

Cost-aware conversation and observability

The conversation implementation bounds recent message history and checks subscription or credit access before AI calls. A usage logger records tokens, audio duration, model, feature, and estimated cost, making consumption visible instead of treating the AI service as an unmetered dependency.

Commerce linked to application access

WooCommerce order and payment events connect purchases to subscription access. The integration also includes expiry and revocation handling, keeping learning permissions tied to the account lifecycle.

Outcome

Launched a live platform that brings structured learning, AI feedback, and career information into one guided experience. The screenshot shows the current public site; features and presentation may evolve beyond the original project scope.

What the platform includes

  • JLPT N5–N1 curriculum organized by level, unit, and skill
  • AI-assisted reading evaluation and speaking practice
  • Learner accounts and progress tracking
  • Career and job information with an LPK application flow
  • Mobile-first, responsive interface
  • Scenario-based AI conversation and interview practice
  • WooCommerce-linked subscriptions and credit controls
  • AI usage records for tokens, audio, and estimated cost

Technology

WordPress · WooCommerce · PHP · MySQL · JavaScript · REST API · Speech-to-text · AI integration

Visit the live site

Related reading

Back to selected work

Have a role or project in mind?

For opportunities, collaborations, and conversations about software.

Get in touch