Logo

Uwords: Language Learning App Based on Your Own Speech

Industry:
Education
Timeline:
8 months
Type:
Mobile ApplicationMini App
Platform:
iOS / Android / Telegram

A tool for personalized foreign vocabulary learning, built around AI speech analysis and gamified training

[client]

/01

The product was developed as an alternative to standard products where learning follows a universal scenario.

At the core of the concept is active speech analysis of the user and the formation of a personalized vocabulary based on it.

The goal is to create a system where every learning element is personalized: from word selection to exercise structure, and game mechanics support regularity and engagement, keeping the focus on results.

[key features]

/02

Speech Analysis and Vocabulary Building

The AI module recognizes the user's speech, identifies active vocabulary and builds an individual learning path.

Adaptive Word Retention

Each word goes through a series of exercises with varying cognitive load. The sequence depends on errors and learning dynamics.

Lexical Wave

The main learning track where tasks alternate in difficulty and type. The user moves along an individual learning curve.

Gamified Training Modes

Game formats with progression, XP, and error tracking boost engagement and material retention.

Regularity and Motivation System

Streak system, levels, and rewards are built into learning. They maintain rhythm and form habits.

UWORDS application

[product hypothesis verification process]

/03
1 User interviews
2 Competitor analysis
3 UI prototyping and UX tests
4 A/B tests of progressions and mechanics
5 MVP with speech recognition

[development process]

/04
1 Hypothesis setting and product model formation
2 Designing workouts, progressions, and game scenarios
3 Creating interface prototypes
4 MVP with speech recognition and adaptive task wave
5 Testing and user behavior analysis
UWORDS application

[team]

/05
#Specialist
1 Product Manager
2 UX/UI Designer
3 Backend Engineer
4 Frontend Engineer
5 AI Engineer
6 QA Specialist
[tech stack]
/06
React Native
React Native
React
React
TypeScript
TypeScript
GraphQL
GraphQL
OpenAI API
OpenAI API
Telegram API
Telegram API
Python
Python
FastAPI
FastAPI
PostgreSQL
PostgreSQL
Redis
Redis
RabbitMQ
RabbitMQ
Docker
Docker

[technical challenges and solutions]

/07
01
Challenge

Implement three-level permissions (role, department, entity type)

Solution

Custom ACL module with caching and flexible configuration

02
Challenge

Improve relevance of AI suggestions and reactions

Solution

Building an intermediate layer with enriched context — routes, roles, statuses, task parameters

03
Challenge

Prevent stuck tasks and implicit blockers

Solution

AI generates text comments explaining the reasons and suggests actions (e.g., 'To continue, you need to...')

[results]

/08
  • return the next day
    > 70 %
  • Average session duration
    11 minutes
  • used voice analysis at least once
    83 %
  • fully learned words
    60 %
  • Subscribed on day 5
    18 %