BentoCloud

A fully-managed platform designed for building and operating AI applications.

Overview

BentoCloud is a developer platform designed for AI teams to build, deploy, and manage AI applications efficiently. As the solo designer, I led the UX redesign to improve user onboarding and streamline day-to-day workflows. This involved close collaboration with the founding team and engineers, and continual iteration based on user feedback.

Timeline

April 2024 - Ongoing

User Need

Understand platform features and effectively deploy AI services.

Business Need

Reduce Time to Value

Boost Onboarding Completion Rate

Team

Xijia Zhang

Chaoyu Yang (PM)

Jinyang Liu (UI Engineer)

Sean Sheng (Lead Engineer)

Impact

The redesigned BentoCloud platform resulted in an 8x increase in onboarding completion rates and a 16% improvement in the NPS score. These metrics highlighted the success of the redesign in meeting user needs and business goals.

Context

Why we need redesign?

BentoCloud’s initial prototype, developed quickly by the founding team and engineers, was functionally sound but posed significant usability challenges. Users, primarily AI engineers and managers, found the platform difficult to navigate and overwhelming, which impacted their ability to effectively deploy AI services.

Research

Who are the Users?

The primary users of BentoCloud are AI teams within mid to large-sized enterprises. These teams are often composed of data scientists, machine learning engineers, and AI managers who are responsible for developing, deploying, and managing AI applications within their organizations.

Research Insights

1. Existing Users

The current interfaces were overwhelming and hard to find information for deploying AI services.

2. New Users

Struggled with understanding the platform's features and get onboard quickly.

Problem Statement

How might we help AI teams onboard BentoCloud faster and deploy AI applications more efficiently?

Design Process

The redesign was approached iteratively, with continuous collaboration between myself, the PM, and the engineering team. Frequent discussions and feedback loops ensured alignment and allowed for quick refinements.

Feature Planning

Problem1

Deploy a Bento - How to deploy AI apps more efficiently?

It’s an entry point to Orby experience. It gives the user suggestions for the new contracts in the queue in Google Drive. The users can also search in the list, expand to view more details, and manually add suggestions to the list.

Old Flow

New Flow

Streamlined User Flow

Bento List

Bento Detail

Provide the latest version of each Bento and only show the most important attributes.

Show detailed information of the Bento and allow user to switch between different versions

Iterations

During the sync-up meeting, we found that the function of filtering by label is hindered by the current design that only displays the latest version. And it makes it hard for AI developers to find the Bento in old versions, but if we show all the versions it will also raise another problem that hinders users from quickly getting an overview of all the Bentos. To figure out what is the best solution, I sat down with the PM and Head of Engineer to understand the needs and scenarios when users use this page. Through our discussion, I felt a sense of clarity and realized that we had two distinct user groups: