Deep dive
Read the full case study
The technical details, the tradeoffs, and the results — written down properly.
Read case study Back to projects
AI & Machine LearningEducationongoing
NihonGO! AI Japanese Tutor
AI Japanese tutoring platform with 3D avatar conversation, real-time error correction and FSRS spaced repetition across six languages.
2026 12 months 1 team members
449 passing
tests
23 with RLS
tables
4 LLM backends
providers
6 UI locales
languages
Overview
NihonGO! pairs learners with 3D avatar characters for one-on-one Japanese conversation practice. A streaming tutor replies in pure Japanese while a separate analysis endpoint returns corrections, vocabulary and grammar notes. The product spans a Flutter app (mobile, web, desktop), a Hono/TypeScript backend on Supabase, and four LLM providers behind one interface — with an internal security audit completed before store submission.
The challenge
Japanese tutoring feedback has to arrive while the learner is still engaged — streaming replies, spoken output, and corrections — on an architecture that also had to work offline-first and survive a security review of its payment paths.
Our solution
Built a streaming conversation loop with a separate analysis endpoint for corrections, FSRS scheduling for vocabulary, VOICEVOX text-to-speech and Whisper speech-to-text, and an LLM provider abstraction with Redis caching and fallback. An internal audit hardened payments, CORS, rate limits and CI before store submission.
Key features
Avatar conversation with lip-sync and voice output
Streaming Japanese chat with a dedicated corrections/analysis endpoint
FSRS spaced-repetition cards built from conversation mistakes
Shadowing and JLPT practice modules
Offline-first local database with sync queue
Streak tracking computed in the learner timezone
Results & impact
263 backend tests passing (from 25 failing at audit start)
186 Flutter tests passing (from 7 failing)
23 PostgreSQL tables with row-level security on all of them
4 LLM providers behind one interface with Redis cache and fallback
Six UI languages: EN, JA, VI, ZH, KO, ES
Payments, CORS and rate limits server-verified after security audit
Related Projects
Explore more projects similar to this one
AI & Machine Learning
2026
Kaiwa Coach
Kaiwa Coach scores spoken Japanese entirely on-device. A Rust engine (called through Dart FFI) runs voice activity detection, YIN pitch tracking, mora segmentation and DTW alignment against native reference contours. Flutter handles the interface, encrypted local storage and sync, while Firebase acts as a control plane only — no audio ever leaves the phone.
FlutterDartRust+6
View Project
AI & Machine Learning
2025
SentinelVision Edge
SentinelVision Edge turns low-cost Raspberry Pi cameras into an AI security fleet. Node agents run YOLO person and face detection at the edge, publish occupancy and dwell-time signals over MQTT, and stream video through MediaMTX. A FastAPI backend with PostgreSQL, Redis and Celery manages cameras, areas, recordings and roles, with a Vue 3 dashboard for real-time status.
Raspberry PiYOLOUltralytics+10
View Project
AI & Machine Learning
2025
VietProStocks
VietProStocks combines Vietnamese market data, technical analysis and a retrieval-augmented LLM into a research assistant. The RAG stack — MiniLM embeddings, a FAISS index and a local llama.cpp model — is designed to run entirely on a Raspberry Pi 4 8GB, with an OpenAI-compatible backend as an escape hatch when more compute is available.
PythonFastAPISQLAlchemy+9
View Project
Start Your Project
Tell us about your project and we'll get back to you within 24 hours.
Reference Project:
NihonGO! AI Japanese Tutor