Overview
A text generation service, hosting the latest large language models (LLMs) for inference. The key objective was to create a user-friendly platform that seamlessly integrates into users' existing workflows while leveraging a robust and flexible design system originally developed for image generation use cases.
As the lead UX Designer on this project, I worked closely with engineering and senior stakeholders to improve product functionality and user experience. My responsibilities included coordinating across teams, conducting user research testing, and developing a flexible design system. These efforts were crucial in facilitating significant growth in self-service customers and ensuring the platform handled high traffic volumes with reliability.
Challenges
Seamless Integration: Ensuring the service could be easily integrated into diverse user workflows.
User Experience: Creating an intuitive interface that catered to both novice and experienced users.
Scalability and Reliability: Designing a system capable of handling over 18 billion tokens per day.
Feedback Incorporation: Continuously integrating user feedback to reduce friction in the first-time user experience (FTUX) and application integration processes.
Smart defaults
An intuitive interface caters to both novice and experienced users, with no fuss defaults and advanced parameter settings.
Seamless integration
Streamlined from testing to deployment: Ready-to-use code snippets and in-app API parameter viewing.
Minimal design
A minimal design with a high-contrast color scheme and subtle animations for natural human-AI chat interactions.
Multi-modal
Newly released Open Source (OS) models feature multi-modal capabilities, allowing users to generate text from image input.
Approach
User Research and Testing:
Conducted extensive user research to understand pain points and requirements.
Organized local hackathons with a diverse group of users to gather actionable insights on usability.
Created user personas and journey maps to guide design decisions.
Design System Development:
Adapted the existing design system to support text generation use cases while maintaining its flexibility for future enhancements.
Developed a component library that included API parameter forms, and code snippet generators.
Ensured consistency and scalability across the platform by standardizing design elements.
Seamless Workflow Integration:
Designed an interface that allowed users to test models responses with low latency and cost.
Created interactive model demos that provided users with API parameters and ready-to-use code snippets.
Integrated features that allowed users to customize style and fine-tune models directly within their workflows.
Collaboration and Coordination:
Worked closely with engineering to ensure technical feasibility and optimal performance of design solutions.
Engaged with senior stakeholders to align design goals with business objectives.
Iterative Design and Feedback Loop:
Implemented an iterative design process, releasing updates based on user feedback and performance data.
Reduced friction in FTUX by simplifying onboarding processes and providing comprehensive documentation.
Continuously monitored user behavior and satisfaction, making data-driven adjustments to improve the overall experience.
Impact
Growth in Self-Service Customers: Achieved a 278% increase in self-service customers over the past year by effectively incorporating user feedback and reducing integration friction.
High Traffic Handling: Successfully designed a platform capable of processing over 18 billion tokens per day with high reliability.
User Satisfaction: Received positive feedback on the ease of use and seamless integration capabilities of the platform, leading to increased user adoption and retention.
Conclusion
This project highlights the importance of user-centered design in enhancing complex AI services. By prioritizing seamless integration, robust functionality, and continuous feedback incorporation, we were able to create a highly effective and user-friendly text generation solution. The collaborative efforts across teams and a flexible design system were pivotal in achieving significant growth and ensuring the platform's success.
