Inspiration
Most learning tools today either rely on walls of text or feel robotic and impersonal. We asked: “What if tutoring felt like a real classroom?” — where a student could talk to a tutor who thinks, responds, and even has a face and voice.
That’s how MyCogniAI was born — a browser-based tutor powered by AI that you can actually see, hear, and talk to.
What it does
MyCogniAI simulates a lifelike AI tutor experience designed to replicate the feel of talking to a real human teacher. Users can type questions directly into the chat interface, and the tutor generates intelligent responses using Google’s Gemini language model. These responses are then spoken aloud using ElevenLabs’ voice synthesis, while a looping avatar animates to simulate real-time engagement. No logins, no setup — just type and learn instantly. The overall result is an immersive chat experience where learners can interact with an AI tutor that feels surprisingly natural and responsive.
How we built it
On the frontend, we used React and TailwindCSS to build a clean, responsive user interface that adapts beautifully across devices. For the backend, we chose Java Spring Boot, and integrated it with Google Gemini 1.5 Flash via HTTP requests using Unirest. The AI’s spoken responses are generated through ElevenLabs' API, and synchronized with looping MP4 video avatars that represent the tutor visually. We carefully designed each interaction so the avatar only moves when the AI is speaking, creating a seamless and engaging experience.
Challenges we ran into
One of the trickiest parts of building MyCogniAI was handling the integration between the Gemini language model and ElevenLabs. Gemini’s output had to be cleaned and pre-processed so it could be passed smoothly into ElevenLabs for speech synthesis. We also faced a real challenge synchronizing the avatar’s visual feedback with the voice output, making sure the avatar appears idle or animated at the right times. Time constraints were intense — building the entire system (backend, frontend, avatars, voice integration) in less than 48 hours. We also had hardware limitations — My mic stopped working during development, so we pivoted to AI-generated voiceovers for the demo. I had drivers issues so I couldn't use OBS Studio to record the demo and I had to fix that to come up with something. I also ran out of free credits Eleven labs gave and so I couldn't even make enough tests in the video demo i did. But against all limitation, I'm glad I came up with something.
🏆 Accomplishments We’re Proud Of
We’re incredibly proud of how immersive and seamless MyCogniAI feels. Despite the limited time, we successfully simulated a thinking, talking AI tutor directly in the browser — no setup required. We pulled off full integration between Gemini, ElevenLabs, and our custom-built avatar system. All of this was developed from scratch during the hackathon, and the end result is an AI-powered learning tool that feels smooth, human-like, and futuristic.
📚 What We Learned
This project taught us how to build pipelines that combine multiple AI services — from LLMs to TTS engines — into a unified user experience. We gained hands-on experience using Spring Boot to dynamically pass data like Gemini output through headers, managing both audio and metadata in a single HTTP response. We also learned how to design real-time chat UIs that sync with visual media, and more importantly, how to adapt quickly when challenges arise. AI tools not only powered the app — they also helped us overcome roadblocks when it mattered most.
🔮 What’s Next for MyCogniAI
This is just the beginning. We plan to introduce webcam support, so the AI tutor can eventually "see" the learner — bringing us even closer to true two-way immersive sessions. We also want to add voice input, letting users talk naturally instead of typing. Other upcoming features include subject-specific tutoring personalities, learning memory that can remember past conversations, and a full student dashboard that tracks progress and suggests personalized learning goals. We're just getting started.
Thank you for checking out MyCogniAI! We believe this is what the future of learning looks like — and we’re excited to keep building it 🚀
Built With
- elevenlabs
- gemini
- react
- shadcn
- springboot
- tailwind
- typescript
Log in or sign up for Devpost to join the conversation.