Back to other cases
Period
Under 24 hours, 2026
Tools
Figma, OpenAI (Codex), Agora
Industry
EdTech / Language Learning
Team
2 people — myself and a developer
My role
Idea & research
Overview
Research & Insights
Key Design Decisions
Reflections & Learnings
Preply × OpenAI · AI Shadowing · Hackathon –  Preply × OpenAI · AI Shadowing · Hackathon – 

Preply × OpenAI · AI Shadowing Feature

Overview
During our presentation
My teammate and I
Application form fields
We signed up for the event as Team Pepperoni. Any guesses what our favorite pizza is?
Hackathon invite message
The email we got after we were accepted to the event.
The challenge: My team was just two people — me and a developer. We came in with a clear idea from day one. As a Preply learner myself, I kept facing the same problem: not enough speaking practice. My research showed that shadowing is one of the best ways to fix this. So we decided to turn that insight into a working, AI-powered feature — designed, built, and pitched — before time ran out.
What is Preply Hackathon: Preply × OpenAI Hackathon — a 24-hour challenge to find ways of integrating AI into Preply and improve the language-learning experience.
Research & Insights
Research & Insights
Project Goals4 goals
Preply · Discovery
With under 24 hours, I set goals for my field research during networking.
🎯 Project Goals
Understand the biggest barriers to speaking a new language
Validate that speaking is a shared pain, not just my own
Find the specific moments that cause the most friction
Understand what existing solutions offer and find our gap

People are afraid to make mistakes — they freeze when it's time to speak

Field research

People struggle to reproduce the right accent and pronunciation

Field research

People lack the active vocabulary to express what they mean

Field research

Speaking feels more stressful than reading or listening, because a mistake is immediate and public

Field research

Almost everyone I spoke to named speaking as their weakest skill

Field research

Many said they understand far more than they can produce

Field research

People know words in theory but can't recall them in real time

Field research

Without feedback, people repeat the same mistakes and don't know how to improve

Field research

Strong shadowing apps exist and prove the method works

Competitive analysis

But they operate in isolation — disconnected from real lessons or a tutor

Competitive analysis

None personalize practice based on a user's actual conversations

Competitive analysis
Key Design Decisions
Step 1: Session entry

After a tutor session, AI analyzes the conversation and selects relevant phrases for shadowing practice.

Session entry
Design decision 01

We automated spaced repetition to remove the effort of tracking study intervals. The algorithm schedules reviews at the optimal time for better language retention.

Product Impact

This drives Day 1–7 Retention by removing the effort of deciding what to study next.

Spaced repetition system
Preply — design decision 01 detail
Preply — design decision 01 detail
Step 2: Practice — Repeat & get feedback

Student repeats a phrase up to 4 times per set. After each attempt, AI gives targeted feedback on what to improve next.

Practice — repeat and get feedback
Design decision 02

We introduced real-time pronunciation feedback to encourage active speaking practice.

Product Impact

This reduces in-session drop-off by providing immediate corrective feedback.

Real-time pronunciation feedback
Preply — design decision 02 detail
Preply — design decision 02 detail
Step 3: Set summary

After repeating the phrase 4 times, the student sees a summary of their performance across all attempts in that set.

Set summary — progress after 4 attempts
Design decision 03

We built a per-set progress summary with scored metrics — so students can see improvement across attempts and feel motivated to continue.

Product Impact

This increases Session Depth by motivating users to retry and improve their scores.

Word-level accuracy
Preply — design decision 03 detail
Preply — design decision 03 detail
Step 4: Final review — All insights

After all 3 sets, the student gets a full summary — overall progress, patterns across rounds, and a concrete next step.

Final review — all insights
Design decision 04

We created a structured end-of-session review with a single next-step recommendation — so students leave with a clear reason to come back.

Product Impact

This drives long-term retention and LTV by connecting independent practice with tutoring sessions.

Vocabulary building
Preply shadowing — user flow
Reflections & Learnings
Reflection card — on scoping under constraints
Reflection card — on challenging my own idea
Reflection card — on working backwards from the outcome
Reflection card — on presentation
Tanya Rybalchenko

Nice to see you here!
Let's connect via...

...and talk about your product, ideas, or just have a virtual coffee together!

LinkedIn Instagram Dribbble [email protected]