MathWorks Stats & ML Toolbox
At MathWorks, I was part of the MATLAB Machine Learning toolbox team to help MATLAB users perform ML tasks. I collaborated with a UX researcher and a developer for my internship project.
According to usability tests, my design reduces by 47% the comparison time over the current design. The new design will also be developed and published in the Fall 2022.
Role: UX Design intern
Time: 4 months, 2021
Tools: Figma
Team: 1 UX researcher and 1 developer
PROJECT
Machine learning usually requires multiple trials and errors. Thus, users often end up with more than 20 models to select from. However, the current app does not provide an efficient way to compare different models.
I was challenged to design a solution to help users compare and find the best performance models.
Inefficient workflow
It’s impossible to sort models in the Models panel using one metric and inspect the other at the same time. Users need to click back and forth for each model.
Cluttered side panel
When there are many models trained in an app session, the Models panel looks crowded and becomes hard to navigate.
Due to the NDA, I cannot share more details on my design publicly. Please reach out to me if you are interested！
Reflection
Understand ML:
Before I started my internship at MathWorks, I had no prior knowledge of Machine Learning. I spent the first 2 weeks chatting with ML experts, reading articles, and watching tutorials to get familiarized with all ML basic concepts and general workflow. In the end, I even acquired a certificate by completing the intro course for ML at MathWorks.
Work with Constraints:
Unlike school projects where I only need to consider users' needs to craft my design, this project requires my design to be down-to-earth and ready to implement. Often than not, I needed to consider the technical constraints from the engineering side and compromise my "ideal" design to a more realistic one. Thanks to the compromises I made, my design is possible to be developed and published in Fall 2022.
