Inspiration
What it does
How we built it
Challenges we ran into
Accomplishments that we're proud of
What we learnedInspiration
Learning algorithms through text is frustrating. We kept breaking our coding flow to type questions to ChatGPT or search Stack Overflow. We realized: voice is how humans naturally think through problems. Why not make that the primary interface for AI-powered learning? What it does KachowAI is a voice-first AI mentor for learning data structures and algorithms. You speak naturally while coding, and it: Automatically watches your code files in real-time Guides you with Socratic questions (never just gives answers) Responds through natural voice conversation Helps you build intuition, not just memorize solutions Think: pair programming with a patient senior engineer who's available 24/7. How we built it Backend: FastAPI + Google Gemini 2.5 Flash for intelligent reasoning Voice: Web Speech API for input, ElevenLabs API for natural TTS output Real-time sync: Watchdog file monitoring + WebSockets for instant code context updates Frontend: React with glassmorphism UI for a modern, clean experience Teaching logic: Custom Socratic prompting system that guides without spoiling Challenges we ran into Speech synthesis was robotic - Code syntax like "O(N^2)" became "back quote N squared back quote" Solution: Built text sanitization to convert code patterns into natural speech Context management - Balancing file watching performance with comprehensive code understanding Solution: Smart filtering by extension + 5-file rolling window Teaching, not telling - AI naturally wants to give complete answers Solution: Refined prompting system with strict Socratic guidelines
Built With
- cursor
- elevenlabs
Log in or sign up for Devpost to join the conversation.