Fluent AI

Master fluency with Fluent AI
Start a custom conversation, preset conversation, or just have an open conversation
Relying on listening and speaking as you make your way through a conversation without the transcript
Get fluency, clarity, and accuracy feedback on each conversation
See suggestions to improve your word choice
Fluent listens to your conversations and automatically updates your fluency ratings based on how you speak

Inspiration

After spending ~10,000 hours learning and speaking foreign languages, I became frustrated with the lack of AI tools focused on real conversational fluency—especially for Mandarin. Existing apps don't help users build the spontaneous, dynamic fluency needed in real conversations, so I built my own.

What It Does

Fluent is an AI conversation partner that helps users develop true fluency. It:

Provides flexible realtime conversations
Tracks fluency metrics (WPM, response time, pronunciation clarity)
Flags grammar issues, awkward phrasing, and unnatural word choices
Offers targeted feedback and vocab review lists
Adapts learning plans based on user conversations
Measures and displays progress through dynamic scenario-based metrics

Tech Stack

Built with Firebase, OpenAI, React Native, Flask, Google Cloud Store, and NLP libraries (jieba, spaCy, etc.). Integrates speech-to-speech interaction and transcript processing for vocabulary and scoring analysis.

Challenges

So far, designing helpful, non-overwhelming feedback was the hardest part. The next big challenge is pronunciation feedback—currently exploring ML models and relevant datasets for this.

Accomplishments That We're Proud Of

I began this project 6 days ago and built the core MVP. The 2-day hackathon portion accounted for the majority of the technical and UX progress.

Before the hackathon (Monday-Thursday):

Basic UI
Unreliable real-time chatbot
Basic fluency and clarity metrics

During the 2-day hackathon:

Gathered feedback from 3 users (including a Chinese lit Ph.D. student), who completed more than 10 conversations
Incorporated full transcripts and NLP review
Improved reliability of chatbot and fluency/clarity scoring
Introduced new metrics: grammar, word choice, listening comprehension
Built automatic coaching and personalized vocab review
Added flashcard review system

Key Learnings

In conversation, fluency trumps accuracy
Users respond to visible progress on fluency metrics.
Few (if any) tools apply full-stack NLP to drive language acquisition.

Why Me?

10K+ hours studying/speaking second languages
Advanced Chinese Scholar, U.S. Flagship Program; lived in Taiwan 2 years
Advanced HSK Oral Proficiency (e.g. top-rated Mandarin fluency)
Former data scientist (Meta, Capital One)
Grad certificate in AI/NLP from Stanford
Studied Arabic, French, Spanish, and Mandarin

What's Next

The MVP is far enough along to shift focus to validating core hypotheses and finding product-market fit before tackling the bigger technical questions like precise pronunciation scoring at scale.

Product-Market Fit: