Redesigned Bodo.ai to simplify cluster and workspace tasks, boosting UX by 40% through user research and cross-functional collaboration. 

My goal, as the lead UX designer, was to transform the Bodo.ai platform, making it much easier to navigate and perform tasks related to clusters and workspaces. This led to a substantial 40% improvement in user experience. This achievement was the outcome of in-depth research, creative planning, prototype building, and insightful conversations with users. I took charge of developing new ideas, mapping out the process, crafting detailed blueprints, and designing an appealing interface. I collaborated closely with project managers, subject matter experts, and software developers to uphold exceptional UX standards. Additionally, I conducted thorough interviews with 7 users, gathering invaluable feedback to further refine and enhance the Bodo.ai platform. 

Problem
Even Experienced Users Struggle to Complete Key Tasks on the Platform

Before the redesign, Bodo users, even experienced data engineers and scientists, faced challenges navigating the platform to complete essential tasks like configuring clusters, managing workspaces, and selecting the right instance types. The lack of guidance, unclear permissions, and inconsistent navigation led to frequent misconfigurations, accidental deletions, and confusion about costs and performance. These issues resulted in higher support tickets, user frustration, and directly threatened adoption, retention, and overall trust in the platform. 

Design Goal

To improve usability and reduce critical user errors by enhancing the clarity of job status, permissions, and workflows, making it easier for users to confidently manage clusters, configure jobs, and navigate the platform without confusion or unnecessary support. This directly supports improving user adoption, retention, and operational efficiency.

The solution

The Smart Recommendation Engine and Cost Estimator reduce friction and prevent misconfigurations by guiding users through targeted questions and real-time feedback. 

The platform’s complexity makes it difficult even for experienced users to confidently complete essential tasks. To address this, we propose an intelligent assistant that leverages smart UX, product design, and automation to simplify setup. The Smart Cluster Recommendation Engine guides users through a brief workflow, asking a few targeted questions to recommend the optimal machine type, Bodo version, instance configuration, and scaling strategy. Paired with a Cost Estimator with Slider Control, users receive real-time pricing feedback to better understand cost-performance trade-offs. Together, these features reduce friction, prevent misconfigurations, and enable users to make confident, cost-effective decisions. 

Smart recommendation engine simplifies decisions and configuration complexity. 

Cost Estimator with historical usage patterns

Competitive Analysis

Competitors Streamline Data Analysis with Modular Interfaces and Cloud-Based Flexibility

Leading platforms like Databricks and Snowflake stand out for their user-centric design and advanced data-handling capabilities. Databricks offers IPython-style notebooks and automated cluster management, supported by a modular UI and side navigation that simplifies workflow and feature discovery. Snowflake emphasizes cloud-based accessibility, with options for editing cluster properties and a dedicated support menu, empowering users with seamless, scalable data analysis experiences. 

User Research Insights

I performed user interviews and surveys to better understand the users and their requirements. I conducted the evaluation, followed by several usability testing sessions with data scientists using the beta platform with the help of the product manager and chief engineer. In the end, we identified the key areas for improvement in the platform.

Design Iterations

These early ideas were rejected because they didn’t balance ease of use with guidance. The setup wizard was too rigid for advanced users, while the full control dashboard overwhelmed beginners with too many options and no help. Chat-like guided questions leave users confused and prone to mistakes. We needed an approach that guides users, lets them adjust settings, and shows live cost estimates, helping them make better, more confident choices. 

• Added the capability to edit the cluster properties such as name, type, and instance number. 

• Added a permission setting to the cluster to control user access rights.

Usability Testing Feedback (Internal QA)

Conclusions+Lessons Learned

Wrapping Up & What I Learned

The project led to positive outcomes by simplifying the user journey, removing unnecessary steps, and reducing confusion. I explored different design approaches and visual elements to better align with Bodo’s goals. I also practiced explaining my design decisions to the team, which helped get many of my ideas into the final product.

One mistake I made was trying to solve too many problems at once. I learned that it’s better to focus on fixing the core issues first, instead of spreading my efforts too thin. This helped me become more strategic and thoughtful in my design process.

If I had more time, what would I do...

• A/B test new permission UX with 20 beta users

• Instrument cluster deletion flows to track modal interactions

• Measure the drop in error rate from deletion attempts without permission

• Make simple guides or videos for new features.