2026 — Present
HearLoom: AI-powered storytelling for multi-generational connection.
The vision
Enable families to co-create personalized AI stories with custom characters and voices. Bridge the gap between grandparents and grandchildren through storytelling that feels personal, authentic, and deeply connected.
The stack
A full-stack AI application built from first principles. FastAPI backend with serverless orchestration, Next.js frontend with real-time state management, MongoDB for flexible schema, and multi-model AI integration (OpenRouter LLMs, DALL-E image generation, Fish Audio for voice).
Desktop
Mobile
Core features shipped
- Character creation with AI-generated personalized images via DALL-E
- Story generation and refinement using OpenRouter LLMs with custom system prompts
- Multi-language, multi-voice text-to-speech via Fish Audio
- Real-time audio streaming and playback with progress tracking
- JWT-based authentication with Google OAuth integration
- Email sharing via Resend, SMS via Twilio, and social sharing
- Admin dashboard for monitoring usage, errors, and cost tracking
- Comprehensive error tracking and alerting pipeline
Technical highlights
- Scalable backend architecture with async/await patterns for high concurrency
- Multi-model AI orchestration with cost and latency optimization
- Real-time job scheduling for async tasks (APScheduler, Redis-backed)
- S3-based CDN delivery for audio and image assets at scale
- Event-driven cost tracking and usage analytics
- Comprehensive API design with OpenAPI documentation
- Database optimization for fast querying across users, stories, and characters



