NEURAL NETWORK TOOL

 

An engineering team came to me to design a development tool for creating neural networks in digital form as an interface for a new artificial intelligence chipset.

CHALLENGE

How can we make a universal tool for coding neural networks for artificial intelligence?

IMPACT

Created a novel tool that optimized work flow for both engineering and neuroscientists, saving time and increasing understanding. Co-inventor on patent for new interaction method for programming neural networks.

ROLE

UX strategy and art direction. In collaboration with user experience researchers, neuroscientists, engineers, and product managers.

 

     
     

    UX STRATEGY PROCESS

    neoprocess (1 of 1).jpg
     

    Understand the potential users in order to design an intuitive interface.

    • Understand the domain - gained basic understanding of fundamentals and terminology of neuroscience
    • User interviews and observations - identified two key user groups with different work flows and gathered current tools: engineers and scientist
    • Competitive analysis - identified two mental models: 1) code based, 2) neural network based
    • Preliminary design - synthesized data to design one unified user interface experience that addressed both mental models by switching between code and neural network views
    • Participatory design - validated and iterated through co-design with end users leveraging low fidelity prototypes
     

    ART DIRECTION PROCESS

    neoimage.png
     

    Lead all visual aspects of design.

    • Simplified tool design using visual elements
    • Collaborated on the interaction design
    • Prototyped interaction
    • Created assets and specification for implementation
    • Provided user experience support for engineering development
     
     Multiple screens of neural network development workflow

    Multiple screens of neural network development workflow

     
     Visual specification for implementation

    Visual specification for implementation