Our primary inspiration for our app, Synapse, is our goal to help underprivileged individuals, regardless of age, and those with a burning passion to learn about modern technology, such as AI. Synapse is an AI-powered, co-op learning app that matches two users based on their respective skills and drops them into 15-minute timed missions covering topics like Prompt Engineering, AI Safety, AI Privacy, and AI development. One player gets the "Controls" (the interface to input answers), while the other gets the "Intel" (the clues to solve it) — neither can see the other's screen, so they must communicate and teach each other in real-time to beat the clock. After each mission, an AI-powered debrief delivers a personalized Visual Map of Learning with targeted YouTube recommendations. Players can also use The Forge to create entirely new community modules using natural language. The iOS frontend is built using SwiftUI with MVVM architecture. When a user uploads their resume, Gemini parses their skills and interests into a profile string, which our Node.js backend converts into a vector embedding via HuggingFace. MongoDB Atlas Vector Search then searches through the database to find a complementary peer — same interests, different roles — and assigns them as either the Controls or Intel player. Once matched, Gemini analyzes both profiles and dynamically generates an asymmetric 15-minute puzzle in structured JSON. During gameplay, Socket.io handles millisecond-precise clue synchronization between clients, Agora WebRTC powers secure voice chat, and Gemini serves as a live hint system when teams get stuck. When the mission ends, Gemini processes the session performance data and generates a personalized debrief with concepts mastered, areas for improvement, and timestamped YouTube recommendations. The Forge feature lets users prompt Gemini to create entirely new skill modules, which are saved to MongoDB for the community to discover and upvote. The entire project was scaffolded using Google Antigravity as our agentic IDE. Real-time asymmetric state sync between the two mobile clients was a genuinely complex task for us to accomplish. Integrating Agora voice alongside live Gemini evaluation without latency conflicts required careful sequencing, and tuning MongoDB Atlas vector search to return truly complementary matches took significant prompt and embedding iteration. We want to fully launch The Forge as a community-driven ecosystem where users submit natural language prompts to generate new skill modules, which are then upvoted into the official app. Long-term, Synapse has clear pathways into corporate training and EdTech — any domain where two perspectives and real-time communication produce better learning outcomes than solo study. As the community grows, the AI gets better data, matchmaking improves, and the content library expands infinitely without a traditional content team

Built With

Share this project:

Updates