Mentora
Conversational AI tutoring with customizable voice‑interactive tutors in real time.
Visit Live Site↗
An AI-powered educational platform that provides personalized, voice-interactive tutoring experiences tailored to each student's learning style.
Overview
Mentora is a revolutionary AI-powered SaaS platform that transforms online learning into a natural, bidirectional voice experience. Built with Next.js 15 and Supabase, Mentora allows users to architect their own specialized AI tutors—from coding mentors to language coaches. By synchronizing Vapi.ai, ElevenLabs, and Deepgram, the platform achieves sub-second latency in voice-to-voice interaction, providing an immersive educational environment that mirrors real-human tutoring.
Experience Mentora↗My Role
As the Lead Full-Stack Developer, I architected the end-to-end infrastructure of Mentora, orchestrating a complex stack of Vapi.ai, ElevenLabs, and Deepgram to achieve sub-second latency in voice-to-voice tutoring. I engineered a secure, multi-tenant database using Supabase with Row-Level Security and integrated Next.js 15 Server Actions to manage real-time state synchronization across the platform. My work focused on bridging generative AI with human-centric design, implementing a dynamic UI that pairs live transcription with synchronized Lottie animations for a truly immersive learning experience.
Voice-interactive tutoring session interface
Key Features
Mentora features a Real-Time Voice Engine that enables natural, interruption-capable conversations with AI tutors across 10+ academic subjects. The platform includes a Custom Companion Builder allowing users to define specific teaching styles, voices, and topics, complemented by a Live Transcription panel for real-time visual reinforcement. With a robust Session History tracker and a Bookmark System, learners can easily manage their progress and revisit their favorite personalized mentors.
A centralized profile view tracking session data, bookmarked companions, and learning frequency across various academic disciplines.
Key Takeaways
Mentora was a deep dive into Latency Optimization for the modern web; I learned how to minimize the "time-to-first-byte" in voice responses by fine-tuning buffer settings and model parameters. I gained significant experience with Serverless State Management, using Next.js 15 Server Actions and revalidation tags to keep the UI perfectly synced with background database changes without manual refreshes. Additionally, the project reinforced the importance of Accessibility in AI, leading me to implement live transcripts and ARIA-compliant controls for hearing-impaired users.
A modern, high-conversion landing page showcasing the platform's multi-subject capabilities and real-time voice-AI value proposition through a responsive and interactive design.
Mentora bridges the gap between static online courses and expensive private tutoring by providing an always-available, voice-first learning assistant. By architecting a system that handles complex audio processing and generative intelligence in real-time, I have created a scalable solution for personalized education. This project stands as a testament to my ability to build high-performance, AI-integrated SaaS applications that solve real-world accessibility and engagement challenges in the EdTech space.
Certum AI
↗Voice‑powered mock interviews using Hume AI and Gemini for performance feedback.
© 2026 NEMANJA