Based on our research findings and insights from spring semester, we generated storyboards, low-fidelity mockups, and a high-fidelity prototype to arrive at our final solution, TRACE.
Our final product Trace provides a seamless integration of different tools Bloomberg machine learning engineers do, improving on the three components, tracking, documentation, and discoverability.
TRACE
Bloomberg’s New Machine Learning Hub for Simpler Experiment Tracking, Documentation and Discoverability
Information ArchitectureInformation from our research helped us develop the information architecture of the platform. We used the mental models of the ML engineers to steer our design decisions. Then we used a tree test to verify if the information architecture was coherent and efficient. With this information, we were able to start the low fidelity prototyping.Prototypes and Usability StudiesThe information architecture was used as the starting point for building out low fidelity prototypes. These included features that were brainstormed and deemed valuable from our research on Bloomberg’s machine learning workflows. Using usability tests, the features and layouts were iterated on until the final high fidelity prototype was created. The usability studies allowed us to understand how the user is able to navigate through various flows by measuring both time and overall comprehension.
Future StateThe future state would incorporate the core components of our solution (Daily ML Brew, tracking through a comprehensive runs page, as well as a centralized documentation hub).
The all-in-one solution provides a centralized platform for Bloomberg engineers to collaborate more effectively and efficiently.