Model Routing
A Node.js API selected models and parameters by task, subscription level, and response mode.
Product engineering case study
Geppetto
A cross-platform AI character conversation experience designed and shipped across mobile and web, combining expressive product design with a full-stack AI platform.
Explore the case study
Portfolio presentation
Public chat is no longer available
Public chat is no longer available
- Flutter
- Node.js
- Firebase + MongoDB
- iOS, Android + Web
The product
Turning character discovery into an ongoing conversation
Geppetto brought character browsing, AI-assisted creation, persistent conversations, and account management into one focused Flutter experience. The same product foundation supported native mobile releases and a responsive web app.
Character Creation
Guided tools transformed a simple idea into a detailed, ready-to-chat character profile.
Conversation Memory
Stored context and summaries helped conversations retain continuity over time.
Cross-platform UI
One Flutter product system adapted navigation and layouts for mobile and web.
Scene Generation
Optional generated visuals added atmosphere and context to role-play conversations.
Engineering depth
Built beyond the chat interface
Geppetto combined consumer product work with the operational systems required to run an AI application across platforms.
Memory Pipeline
MongoDB-backed messages, summaries, and retrieval kept long-running chats useful without sending unlimited history.
Safety Systems
Moderation, reporting, usage controls, and account tools were integrated into the product workflow.
Subscriptions
Stripe, RevenueCat, App Store, and Play billing supported access rules and subscription lifecycle handling.
System architecture
A full product stack from client to model
The architecture separated the customer experience, product API, and external services so each layer could evolve independently.
01 / Client layer
Flutter Applications
Responsive UI, Firebase authentication, chat state, settings, and native app-store packaging.
02 / Product layer
Node.js API
Conversation workflows, permissions, usage accounting, moderation, WebSockets, and billing coordination.
03 / Service layer
AI + Data Services
Model providers, image generation, MongoDB, Firebase storage, analytics, and payment platforms.
The experience
Designed for discovery and conversation
The product moved from browsing into character details, active chats, and AI-assisted creation through a consistent mobile interface.
A future direction
The underlying platform could be adapted for educational AI simulations that let self-directed learners explore history, literature, and complex ideas through guided conversation.