DEVELOPMENT • GENERATIVE AI • DESIGN • DEVELOPMENT • GENERATIVE AI • DESIGN • DEVELOPMENT • GENERATIVE AI • DESIGN
Challenge
Develop a scalable language learning platform that connects students with native-speaking tutors through personalized matches based on learning goals and communication styles.
Project
We built a custom matching engine on a scalable, container-based infrastructure using AWS EKS and Kubernetes to support the growing demand for video calls without compromising performance.
Scalability was validated with K6 load testing, simulating over 3,000 concurrent users.
AI language coach
By leveraging OpenAI’s Realtime API and Whisper, we developed an agent that processes student speech and generates adaptive responses based on users’ language proficiency.
Students talk to an AI avatar that can maintain a natural conversation flow and use appropriate vocabulary.
High-quality interactions
To ensure session quality, our AI system analyzes video and audio recordings:
- Google Gemini (2.0 Flash) handles visual behavior recognition for teacher conduct compliance.
- Whisper performs speech analysis, tracking words per minute and speaking speed during sessions.
Achievements
Avatar that maintains
Natural conversations
personalized feedback
based on recordings
30,000
registered users
30,000
registered users
+1,000
daily users
+1,000
daily users
20%
annual growth
An immersive, cultural approach to language fluency
Conversifi is a virtual language exchange platform that connects university students across the globe. Through real-time video conversations with native speakers, learners build confidence, cultural awareness, and speaking fluency.
Get in touch!