Design to Code Using a Custom Design System with AI

This talk is scheduled for Nov 21, 16:15
The recording will be published after editing. Multipass and Full ticket holders have early access.
Bookmark
Rate this content

This talk explores how we built an AI-powered system that transforms Figma designs into production-ready React code using Razorpay’s custom Design System. Learn how we solved the problem of brand inconsistency in generic AI tools and created a solution that understands our unique design language, enabling faster development without compromising on quality.

This talk has been presented at React Summit US 2025, check out the latest edition of this React Conference.

FAQ

A design system is a collection of reusable components, patterns, and guidelines that help teams build consistent, high-quality UI efficiently.

Blade is a design system developed at Razorpay to ensure a consistent look and feel across all products, featuring over 65 reusable components and patterns.

While AI can rapidly generate UI code, it cannot replace the consistency, reusability, and accessibility provided by a design system like Blade.

AI can work with design systems like Blade to quickly generate UI code, enhancing productivity by combining the strengths of both AI and design systems.

Blade's MCP (Model Context Protocol) server connects AI models to different data sources and tools, facilitating tasks like converting Figma designs to code.

Blade uses its AI server to interact with Figma and OpenAI, gathering design context and generating UI code based on Figma designs.

Blade provides various tools such as get-blade-component-docs for documentation and get-figma-to-code for converting Figma designs to code.

High-quality documentation is crucial as it helps AI understand how to effectively utilize design system components, improving the accuracy of generated code.

Challenges included transitioning from a Figma plugin to an MCP, measuring generation accuracy, and integrating with other coding tools.

The key takeaway is to maintain concise, accurate documentation, embrace learning, automate mundane tasks, and collaborate with smart, passionate individuals.

Chaitanya Deorukhkar
Chaitanya Deorukhkar
21 Nov, 2025
Video transcription, chapters and summary will be available after the recording is published.

Check out more articles and videos

We constantly think of articles and videos that might spark Git people interest / skill us up or help building a stellar career

Powering Cody Coding Assistant Using LLMs
C3 Dev Festival 2024C3 Dev Festival 2024
29 min
Powering Cody Coding Assistant Using LLMs
This Talk explores the world of coding assistants powered by language models (LLMs) and their use cases in software development. It delves into challenges such as understanding big code and developing models for context in LLMs. The importance of ranking and code context is discussed, along with the use of weak supervision signals and fine-tuning models for code completion. The Talk also touches on the evaluation of models and the future trends in code AI, including automation and the role of tasks, programming languages, and code context.