← back to work

AI · MOBILE · PRODUCT · FULL-STACK

SpeakEasy

AI language-learning app

Three SpeakEasy app screens showing interest selection, AI discussion, and article learning

Role

Product strategy, UX, mobile app build

Stack

Expo, React Native, Node.js, Supabase, OpenAI

Validation

16-user cohort + expert ESL feedback

Codebase

~12,300 lines across mobile and backend

01 · THE WIDE PROBLEM

Billions try. Few feel fluent.

More than two billion people are learning a second or third language, but many still struggle to use it confidently in real moments.

I can read the words, but when it's time to talk, my head goes blank.
Duolingo taught me to say - I'm a tired doctor.
When would I ever say that?
Duolingo app icon
Babbel app icon
Rosetta Stone app icon
Memrise app icon
Pimsleur app icon
Busuu app icon
HelloTalk app icon
Mondly app icon
Lingopie app icon

Popular apps studied

Insights

01
Context beats structure

People learn best when language shows up in a context they understand, not when the curriculum says it's time.

02
Safety builds fluency

Fluency grows from comfort, not perfection. Fear of embarrassment was the single most repeated blocker.

03
Meaning travels through culture

Culture is inseparable from language. Apps teach correctness; people crave meaning.

04
Immersion, not lessons

Living where the language is spoken · Conversing regularly with native speakers · Integrating the language into daily routines · Watching shows in the target language

02 · THE IDEA SPACE

Turning research into nine ideas

The research gave us three bets to test: practice should come from real context, content should be personally meaningful, and speaking should feel safe before it feels polished.

Direction 01

How might we make practice come from a learner's real world?

The strongest learning moments came from context learners already understood: the places they moved through, the media they consumed, and the routines they repeated.

01
Browser-to-lesson companion

turn everyday browsing into short lessons, which became the LangLearn prototype

02
Environment immersion

make café chatter, signs, and videos feel like practice material, later explored through Verba

03
Scenario rehearsal

let learners practice real situations like ordering food or making calls before they happen

Direction 02

How might we teach through content people already care about?

Learners did not want more abstract drills. They wanted language attached to culture, interests, and stories that made the words worth remembering.

01
Story capsules

small lessons around idioms, humor, and customs from media learners already watch

02
Word lens

show how one word changes meaning across articles, videos, and everyday contexts

03
Interest-based feed

let learners choose topics first, which became the backbone of SpeakEasy's article feed

Direction 03

How might we make speaking practice feel low-stakes but real?

Almost every learner described the same gap: they could recognize words, but froze when they had to speak. The product needed rehearsal without embarrassment.

01
Wearable companion

surface small prompts during real-world moments, an early path we chose not to pursue

02
Voice AI buddy

natural conversation with gentle feedback, first tested in Verba and later carried into SpeakEasy

03
Article discussion

turn something the learner just read into a focused spoken conversation

03 · TWO PROTOTYPES

LangLearn and Verba, built and tested

Concept A
LangLearn cover still

LangLearn

micro-immersive daily learning

A Chrome extension that pulls real vocabulary and phrases from your everyday browsing, then turns them into personalized lessons on mobile.

explore prototype →
Concept B
Verba cover still

Verba

an AR language companion

AR smart-glasses that see what you see and chat with you about it in your target language, like a friend beside you, then compile learning materials from your day.

explore prototype →
04 · THE PIVOT

Meet the 4%

After testing LangLearn and Verba, the direction got clearer. People liked practice that came from things they already consumed, but the product had to live somewhere familiar, private, and easy to return to.

What the prototypes taught us

Real content worked

Users responded to practice built around articles, videos, and daily browsing because it already had context.

Mobile felt more reachable

The browser extension felt useful but too heavy. People expected learning to fit into phone-sized moments.

AR was too early

Verba made the interaction feel alive, but smart glasses were not something most people owned or trusted yet.

Voice practice mattered

The buddy-style conversation was the part users kept coming back to: low-stakes, personal, and close to real speaking.

Pivot

A smaller,more believablehabit.

real content+mobile feed+voice buddy

Content first, practice second, no new hardware or classroom schedule required.

05 · THE PRODUCT

A feed you learn inside

The final app became a simple learning loop: choose something worth reading, get help when stuck, then practice speaking about it.

Content feed

Interests shape the feed.

Learners start with topics they care about, so every article has context before the lesson begins.

Content feed feature annotation

Vocabulary support

Help stays in the moment.

A tap brings up pronunciation and native-language meaning without leaving the article.

Vocabulary support feature annotation

Voice discussion

Reading turns into speaking.

The AI buddy turns the article into one focused, low-stakes conversation.

Voice discussion feature annotation

System design

SpeakEasy system design diagram

Download now

Try SpeakEasy on your phone.

App Store

Download SpeakEasy on the App Store
SpeakEasy App Store TestFlight QR code

Play Store

Get SpeakEasy on Google Play
SpeakEasy Android access form QR code