Problem:
The startup was working on an innovative chip and needed to create firmware and supplemental software for optimizing it. In addition to the firmware software, the investment needed to find a blue ocean opening in the software field for new machine learning software to enhance the chip’s presence on the market.
Competitive Analysis:
In the initial stages, outlining and breaking down features of competitors in order to consolidate the demographic of users and new market features.
UX Research:
Interviews with a small sample set of seasoned machine learning experts in various fields helped to discern the most important features and goals of software users. In the research, we needed to differentiate between the type of machine learning and their workflows. Deconstructing software ranging from Lobe to Open Source, such ML along with taking Fast.ai courses for differentiating the blue ocean factors for the startup.
Process & Summary: Upon further explorations, conversational design was decided upon to assist the user in their work flow. The design system and conversational scenarios for creating components helped to solidify the learning system in each iteration. Most of the machine learning platforms only use the terminal for their workflows, but allowing for conversational design to assist in optimizing their process as the demographic for the software expands.
The premise of the application allows for conversational design chat to learn the users’ needs with unsupervised learning. Creating components for the chat system and for the dynamic interactivity was the primary challenge as the use cases were hypothetical, so an initial build for one persona was created for understanding the needs of the software.
Creating interactive designs for the stakeholders as well as iterative user research assisted in the lean agile process. Most of the highly interactive elements were handled by conversational chat and settings. Prototyping the machine learning tooling was an intermediary function while the system learned the interactions in order to discern which tools should reside outside of the system for the user to have as static elements.
Prototype: https://xd.adobe.com/view/dedda47e-7689-4714-8ea1-91fba35eb5ad-8b8e/?fullscreen
