Skip to content
Nirmaan AI

Nirmaan AI

NirmaanAI: AI-Powered Multimodal UI Generator for eCommerce — From Prompt to Design in Minutes.

1.4K
Built at Synapses '25
Winner

Created on 13th April 2025

Nirmaan AI

Nirmaan AI

NirmaanAI: AI-Powered Multimodal UI Generator for eCommerce — From Prompt to Design in Minutes.

The problem Nirmaan AI solves

Creating a landing page or product showcase that is aligned with brand identity is traditionally slow, design-heavy, and requires design + frontend collaboration. NirmaanAI makes this instant.

With just a text prompt, brand colors, font choices, and product image, it generates a production-ready React UI using TailwindCSS. Whether you're launching a new product or A/B testing styles, it cuts down your iteration time from days to minutes.

Use Cases:

  • eCommerce teams launching fast product landing pages
  • Designers & founders prototyping brand-aligned pages
  • Hackers or indie devs building polished frontends rapidly

Challenges we ran into

  • Multimodal fusion logic: Merging user prompt with brand styles (colors, fonts, image) and formatting them into a natural language input for the Mixtral model was tricky. We solved this by building a clean formatting function that maps brand tokens into a cohesive design brief.

  • Live Preview + Code Rendering: Integrating Sandpack + Monaco Editor + Tailwind without breaking layout preview took experimentation. We finalized a dynamic bundling approach that previews code in real-time while keeping performance in check.

  • Image Uploads + Injection: Ensuring uploaded images were hosted instantly (Cloudinary) and referenced inside JSX code output was a challenge. We solved this by sending the image URL with the prompt and letting AI place it within the layout (e.g., inside hero/banner).

  • Rate limits on Groq API: We cached and debounced requests to avoid rapid-fire errors from users regenerating designs frequently.

  • Tailwind CSS issues in JSX parsing: We had to strip comments, wrap elements in fragments, and sanitize classnames to make the output render live without crashing.

Tracks Applied (1)

E2E Networks Track

NirmaanAI aligns perfectly with E2E Networks’ focus on empowering AI and cloud-native innovation. Our project leverages ...Read More
E2E Networks

E2E Networks

Discussion

Builders also viewed

See more projects on Devfolio