LinguaChat

Inspiration

We’ve always been passionate about languages and the power they hold to connect people from different cultures. However, traditional language learning platforms sometimes feel too rigid or lack personalization. With the rise of powerful Large Language Models (LLMs), we saw an opportunity to create a more immersive, conversational experience—one that guides users toward fluency in a way that’s engaging and tailored to their individual needs.

What it does

LinguaChat is a language learning app that focuses on natural, conversation-driven practice. Instead of memorizing vocab lists or following scripted lessons, users can chat in real-time with a responsive AI tutor. The AI recognizes each user’s language level and personal interests, ensuring that each conversation is relevant, educational, and fun. The app also keeps track of users' progress, offering insights into their improvement over time.

How we built it

React Native – We chose React Native to build a cross-platform mobile app. Its component-based architecture let us quickly develop a clean and intuitive UI.

Cohere – We integrated Cohere’s Large Language Model to handle the bulk of the conversational AI tasks. It processes user input and generates contextually relevant replies, creating the “chat tutor” experience.

Firebase – For user authentication, it is scalable in the future.

Flask – We used a lightweight Flask server to handle backend logic and coordinate requests to the LLM.

Challenges we ran into

AI Prompt Engineering: Crafting prompts that would yield helpful, user-friendly, and contextually accurate responses from the model was a constant learning process.
Context Handling: We needed to maintain conversational context so the AI could remember user preferences and past conversation topics. We had to carefully manage the payload sent to Cohere’s API to avoid exceeding token limits. ## Accomplishments that we're proud of
Cross-Platform Deployment: Building a single codebase with React Native that runs on both iOS and Android reduced development time and complexity.
Scalable Architecture: With Firebase and Flask, our backend infrastructure can scale to support more users without compromising on performance. ## What we learned
LLM Integration: We gained deep insights into how to optimize prompt structure and maintain context, which is crucial for a fluid user experience.
Mobile UI/UX: Designing an interactive, real-time chat interface on mobile pushed us to explore various strategies for better performance and usability.
Team Collaboration: Coordinating across the frontend, backend, and AI model taught us how to communicate effectively in a fast-paced environment.

What's next for LinguaChat
Gamification: Incorporate badges, achievement milestones, and weekly language challenges to make learning even more engaging.
Offline Support: Provide local phrase libraries and lessons for areas with limited connectivity.
Multi-Language Expansion: Expand beyond our initial language set, allowing more people worldwide to benefit from conversation-based learning.
Voice Interaction: Integrate speech recognition and text-to-speech features for an even more immersive language practice experience.