AI-Powered Language Learning Platform — Vocab App

The challenge

Vocab is our own product — we wrote the brief, built it, and run it. Traditional vocabulary learning suffers from rote memorization without meaningful associations, suboptimal review timing that wastes time or lets words slip, low engagement without gamification and analytics, and the technical complexity of running real-time quizzes, spaced-repetition logic, and instant AI mnemonic generation simultaneously — all while managing freemium vs premium access tiers.

The solution

We built the whole ecosystem in-house: a React Native mobile app with a high-performance Python (FastAPI) backend. An AI learning engine (Groq) generates memorable mnemonics and rich word intelligence on demand. A Spaced Repetition System implements the proven SM-2 algorithm with a 6-point recall scale to surface words precisely when they're most likely to be forgotten. A multi-mode quiz system (definition match, synonyms, fill-in-the-blank, speed rounds) is enhanced with streaks, heatmaps, and leaderboards. LibreTranslate provides multi-language support with a privacy-focused fallback, and Polar.sh handles webhook-driven checkout and tier-based access. The stack runs on FastAPI, PostgreSQL + AsyncPG, Celery + Redis for background jobs, and Upstash Redis caching, deployed across auto-scaling serverless environments, secured with Google OAuth 2.0, JWT, bcrypt, and Casbin RBAC.

The impact

Because we own the product, we own every trade-off in it: SRS scheduling, AI cost per mnemonic, the freemium-to-premium split, and the mobile UX. Automated tracking optimizes review scheduling without manual intervention, and the React Native + NativeWind UI ships one codebase to iOS and Android. The architecture is built to scale to additional language pairs and higher load. Being our own product also makes it our most inspectable build — we can walk a prospect through the code, the schema, and the decisions we would make differently.

Technologies used