Overview
Selected Work

Nexa AI

https://nexa.ai/

Nexa AI provides industry-leading tiny AI models, local inference frameworks, and end-to-end, on-device AI solutions for developers and enterprises. Nexa AI is building the future of AI where intelligence is efficient, private, and everywhere, reducing reliance on costly infrastructure.

Nexa AI has formed strategic partnerships with industry leaders such as AMD and Lenovo to advance on-device AI capabilities. The lightweight, energy-efficient models are designed to deliver speed and reliability, performing well even in offline or limited-resource environments.

My Role

Designed responsive websites for desktop, tablet, and mobile devices that ensured seamless cross-platform display, leading to an innovative layout redesign that refreshed the brand’s identity.

Spearheaded the design of an AI Model Hub and SLM(Small Language Model) Leaderboard web interface and user experience, creating a centralized platform for developers to search, create, and interact with AI model repositories.

Developed a comprehensive design system that promoted collaboration with cross-functional teams and enhanced consistency, efficiency, and scalability across the platform.

Designed a Nexa desktop app for business users, which provides an intuitive and efficient interface for exploring, testing, and deploying a wide range of on-device AI models.

Duration

6 Months

Tools

Figma
FigJam
Linear

Team

Rita L.(Design Lead)
Iris Z.(UIUX)
Alan Z.(PM)
Engineering Team

Selected Works

AI Model Hub

The On-Device Model Hub(Tiny Model Hub) offers developers a centralized platform to search for and download quantized AI models, including Gemma, LLaMa, Qwen, Stable Diffusion, and Whisper. These models are optimized for on-device applications across natural language processing, computer vision, multimodality, and audio processing.

The hub features a comprehensive filtering system, allowing users to select models based on type, file format, parameters, RAM requirements, and file size. Additionally, the Nexa SDK facilitates seamless local deployment of these models with a single line of code. This integrated approach not only simplifies the development process but also fosters a collaborative community for developers and researchers.

Website

I played a key role in the comprehensive redesign of Nexa AI's website, focusing on engaging target users by creating pages that are cohesive, trustworthy, and intuitive.

This included revamping the homepage to effectively showcase our end-to-end on-device AI solutions, updating the SDK download page for streamlined access to tools, and overhauling the blog to highlight the latest research and developments. I also introduced a "Success Stories" section to build trust with companies interested in partnering with us. These enhancements ensure a seamless user experience and strengthen the company’s credibility.

Current version

Nexa App (for business)

The Nexa App, powered by the Nexa AI SDK, provides an intuitive interface for exploring, testing, and deploying a wide range of on-device AI models, including multimodal capabilities like text, vision, and audio processing. Users can not only download models seamlessly but also customize their performance by adjusting advanced parameters such as temperature, max tokens, top-k, and top-p. This flexibility ensures optimal outputs tailored to specific use cases, all within a user-friendly environment.

Design System

Created a design system that uses reusable components and patterns to ensure consistency across products and empower teams to collaborate better and deliver cohesive experiences. Both product and engineering team provided positive feedback on the new design system stating it greatly increased collaboration efficiency.
No image is shown due to Non-Disclosure Agreement(NDA).

SLM(Small Language Model) Leaderboard

The Nexa AI On-Device SLM Leaderboard is a user-friendly platform that allows comparing the performance of various on-device AI models. It provides detailed information on each model's quality, memory usage, response time, and power consumption. The design enables users to customize the displayed data by editing columns and applying filters, so users can focus on the metrics that matter the most. This makes it simple to find the best AI models for any specific needs. The leaderboard utilizes the Nexa SDK's benchmark framework to ensure accurate and consistent performance evaluations.

Next Project