Skip to content
Eszter Tóth

Case study · 2025–

AI-supported language learning - for humans

A language-learning platform that turns authentic content and targeted grammar practice into personalized study experiences.

Role
Design Engineer
Services
Branding, Web development, UI/UX design, Marketing
Language Shadowing logo
Language Shadowing brand identity

The problem

Most language-learning products either reduce learning to repetitive drills or use AI without giving learners enough structure. Learners need material at the right level, clear feedback and repeated exposure to language in context—not simply more generated content.

The goal was to use AI as a learning partner while keeping the learner in control.

Research

I studied how learners use shadowing, translation, vocabulary review and grammar exercises together. Mapping these workflows revealed a fragmented experience: learners often had to find suitable material, create exercises, check their answers and track unfamiliar vocabulary across several different tools.

The main design opportunity was to connect those activities in one focused learning flow.

Solution

I designed and developed a platform that generates personalized language-shadowing videos and grammar-focused translation exercises based on the learner’s language and CEFR level.

Learners can practise authentic sentences, receive contextual corrections, review vocabulary and repeat spoken material at their own pace. Grammar exercises are generated around specific rules and validated so that the content practises the intended structure rather than merely matching the general topic.

The interface brings content generation, video practice, translation, feedback and vocabulary support into a consistent experience across desktop and mobile.

My contribution

As design engineer, I worked across the complete product:

  • Defined the brand identity and visual language
  • Designed the end-to-end learning experience
  • Built the responsive web application and interaction system
  • Developed AI-assisted generation and correction workflows
  • Created reusable patterns for exercises, feedback, progress and vocabulary
  • Refined prompts and validation rules using real generated examples
  • Shaped the product’s positioning and marketing experience

Results

  • A unified workflow connecting shadowing, grammar, translation and vocabulary
  • Grammar-specific generation across multiple languages and CEFR levels
  • More transparent feedback that explains mistakes instead of only assigning scores
  • A reusable exercise system that can support new languages and grammar curricula
  • A human-centred AI experience designed to support deliberate practice rather than replace it