Yaspeak
EdTech language-learning platform; abandoned due to low demand, demo only.
Designed and launched a production-grade EdTech platform from the ground up, balancing scalability, reliability, and development velocity
Key Contributions
- Successfully founded and launched the EdTech platform
- Implemented user matching system to connect learners with appropriate teachers
- Integrated real-time chat and video call functionality
- Set up automated billing system using Stripe for seamless payments
- Founded the company and led product development
Building an EdTech Platform from Zero to Production
Principal Engineer
Challenge
Yaspeak is an all-in-one EdTech platform connecting qualified tutors with students seeking language instruction. The platform required real-time messaging, live video lessons, integrated billing, and dedicated experiences for both tutors and students. High reliability for paid sessions and scalable architecture were non-negotiable requirements.
- Real-time communication was critical for live paid lessons with zero tolerance for failures
- Billing and session management needed to work seamlessly across tutor and student workflows
- The platform required distinct user experiences for tutors and students with shared underlying infrastructure
Solution
I owned the end-to-end system design and implementation, making deliberate technology choices optimized for the platform's specific requirements. The frontend was built with Next.js for rapid iteration and server-side rendering. The backend used .NET Core for performance and long-term maintainability. PostgreSQL provided transactional guarantees for billing and session data. Azure virtual machines offered infrastructure flexibility during the early growth phase, and Stripe handled payments with robust webhook support.
- WebSocket-based real-time messaging for low-latency chat with simple debugging
- Custom WebRTC video solution initially built for full stack control
- Replaced custom video with VideoSDK after load testing revealed stability issues under concurrent sessions
Results
Successful production launch with stable real-time communication
Fully automated billing system processing tutor payments reliably
Scalable architecture providing solid foundation for growth
Clean separation of tutor and student experiences on shared infrastructure
Established deployment pipeline enabling rapid iteration post-launch
Engineering Approach
The most significant technical decision was replacing the custom WebRTC implementation with VideoSDK. Load testing revealed stability issues under multiple simultaneous sessions, and given the business-critical nature of paid live lessons, I prioritized reliability over owning the entire media pipeline. This project required building deep expertise in React performance optimization, Redux state management, TypeScript, and real-time system debugging, all without modern AI-assisted tooling. The result was a production system that balanced technical ambition with pragmatic delivery.