Duddy.io
An AI-enabled No-Code App Builder that compiles visual designs directly into true native Swift, Kotlin, and React code.
What It Is
Duddy is an AI-powered visual app builder for both consumer creators (B2C) and businesses (B2B SaaS). Most no-code platforms in the market generate slow web views wrapped in native shells (like Cordova or basic webview containers), which leads to sluggish mobile performance, rigid animations, and severe platform limitations.
Duddy was built to eliminate this compromise: you design visually on an intuitive canvas with AI assistance, and when you hit export, the platform generates genuine, idiomatic native source code — Swift for iOS, Kotlin for Android, and modern React for web applications.
What It Does
- Visual Canvas with AI Co-Pilot: Users can drag-and-drop components or prompt the AI engine to generate complex layouts, authentication flows, data lists, and state machines automatically.
- Multi-Platform Native Export: Instead of running inside an interpreted wrapper, exports produce complete native project structures ready to open in Xcode, Android Studio, or modern web hosting.
- Custom Compiler & Transpiler Pipeline: Parses the visual canvas representation into an abstract syntax tree (AST) and transpiles it directly to platform-native idioms (SwiftUI/Swift, Jetpack Compose/Kotlin, React JSX).
- Extensible Logic & Backend Connectors: Supports custom API endpoints, databases, authentication, and state management without locking developers into a proprietary ecosystem.
Architecture & My Contribution
As a founding member and core engineer, I led the System Design, Compiler Pipeline, and Backend Architecture:
-
AST & Code Generation Engine:
- Designed the intermediate representation (IR) layer that sits between the visual canvas state and target language generators.
- Built the compiler logic responsible for mapping visual layout structures into performant, clean native Swift and Kotlin code that actual mobile engineers would be comfortable maintaining.
-
AI Layout Synthesis:
- Implemented prompt-to-component workflows using LLMs to translate natural language specifications into valid Duddy UI AST nodes.
- Guarded against hallucinated properties by enforcing strict schema validation before rendering on the canvas.
-
High-Performance Canvas Architecture:
- Optimized state management on the web canvas to handle deep nested layouts, undo/redo stacks, and real-time live preview rendering without frame drops.
-
Business & Growth Involvement:
- Actively involved with the founding team in defining product roadmaps, enterprise onboarding, sales strategies, and identifying real workflow pain points from early adopters.
Skills Used & Developed
- Compiler & AST Engineering: Transforming visual state models into valid Swift and Kotlin ASTs.
- Full-Stack Architecture: Next.js, TypeScript, Node.js, and high-performance frontend state structures.
- AI Product Integration: Prompt engineering, schema validation, and real-time generation pipelines.
- Technical Leadership: Coordinating cross-platform requirements between web, iOS, and Android ecosystems.
Key Learnings & Takeaways
- Never hide behind wrappers when performance matters: The engineering effort required to build a real compiler pipeline is substantially higher than wrapping an iframe, but the resulting app performance and user trust make it the only sustainable choice.
- Keep generated code human-readable: When developers inspect the exported code, it shouldn't look like minified assembly. Generating clean, idiomatic code makes users confident that they own their project without vendor lock-in.
Skills & Technologies