Jan 2023 - May 2023
Xu Lab, Carnegie Mellon University
Research Assistant & Dashboard Designer
Streamlining biomedical research workflows through intuitive dashboard design
Project Overview
While working as a researcher in Professor Min Xu's lab analyzing AI models on biomedical datasets, I identified a critical inefficiency in our workflow. Our team was losing track of previous experiment data, and the process of manually switching between different model configurations, dataset versions, and performance logs was consuming unnecessary time and effort.
To address this challenge, I pitched a proposal for an internal research dashboard aimed at unifying and simplifying access to documentation and model performance analysis. The resulting dashboard significantly accelerated research progress and improved productivity across the team.
My Role
- •UX/UI Designer
- •User Researcher
- •Frontend Developer
- •Data Visualization Specialist
Technologies Used
The Challenge
The Xu Lab faced several challenges in managing their biomedical research workflow:
- 1Experiment Tracking: Researchers struggled to keep track of previous experiments, their parameters, and results
- 2Data Fragmentation: Critical information was scattered across multiple systems, files, and formats
- 3Visualization Limitations: Existing tools lacked effective ways to visualize and compare model performance
- 4Collaboration Barriers: Team members had difficulty sharing insights and building upon each other's work
Design Process
Understanding Researcher Needs
I began by conducting in-depth interviews with lab members to understand their specific pain points and requirements. This research revealed that researchers needed:
- •A centralized repository for experiment configurations and results
- •Visual comparisons of model performance across different experiments
- •Easy access to historical data and the ability to replicate previous experiments
- •Collaborative features to share insights and annotations
Organizing Complex Data
Based on the research findings, I developed a comprehensive information architecture that organized the dashboard into four main sections:
- •Experiment Library: A searchable database of all experiments with filtering capabilities
- •Performance Analytics: Interactive visualizations of model metrics and comparisons
- •Dataset Management: Version control and metadata for research datasets
- •Collaboration Hub: Shared notes, annotations, and discussion threads
Creating an Intuitive Interface
For the visual design, I focused on creating a clean, distraction-free interface that would allow researchers to focus on their data. Key design decisions included:
- •A dark theme to reduce eye strain during long research sessions
- •Consistent color coding for different types of data and metrics
- •Interactive data visualizations with hover states and tooltips for detailed information
- •Responsive layouts that work well on both desktop monitors and lab tablets
The Solution
The final dashboard solution integrated all the research findings and design decisions into a comprehensive tool that transformed how the lab managed their experiments and data.
Experiment Library
Searchable repository of all experiments with detailed metadata, parameters, and results
Interactive Analytics
Dynamic visualizations allowing researchers to compare metrics across multiple experiments
Dataset Management
Version control system for tracking dataset changes and ensuring experiment reproducibility
Collaboration Tools
Shared notes, annotations, and discussion threads to facilitate team collaboration
Outcomes & Impact
The research dashboard had a significant positive impact on the lab's workflow and productivity. By centralizing experiment data and providing intuitive visualization tools, we were able to:
- 1Accelerate Research: Reduced time spent on experiment setup and data retrieval by 40%
- 2Improve Collaboration: Increased knowledge sharing and reduced duplicate work across the team
- 3Enhance Analysis: Enabled more complex comparisons between experiments, leading to new insights
- 4Ensure Reproducibility: Created a reliable system for experiment documentation and replication
Reduction in time spent on experiment setup
Increase in experiments run per week
Experiment documentation compliance
New research insights discovered
Feedback from the Team
"The dashboard has completely transformed how we approach our experiments. Being able to quickly compare results across different model configurations has accelerated our research significantly. This tool has become an essential part of our daily workflow."
Prof. Min Xu
Lab Director
"As a researcher, I used to spend hours trying to track down specific experiment configurations and results. Now I can find everything in seconds and focus on actual analysis instead of administrative tasks. The visualization tools are particularly helpful for identifying patterns."
Xiang Jiaong
Senior Researcher
Lessons Learned
Research-Driven Design
Taking the time to thoroughly understand researchers' workflows and pain points was crucial to creating a solution that genuinely improved their productivity.
Data Visualization Complexity
Finding the right balance between comprehensive data display and visual clarity required multiple iterations and continuous user feedback.
Technical Integration
Connecting the dashboard to existing research systems required deeper technical understanding than initially anticipated, highlighting the importance of cross-disciplinary knowledge.
Explore My Other CMU Projects
Check out my work with CyLab and the Abstract Lab at Carnegie Mellon.