Problem statement

Gravity needed a new way of shipping the machine learning solution and how we managed to configure our state-of-the-art personalization engine.

Project Goals

The goal is to create an interface, wich is “A bit visual programing, a bit a mind map a tool, a kind of conception modelling, and pinch of visual sales - at the same time.

  • 1

    Help integration and data mining teams work more efficiently and help them to streamline their operations.

  • 2

    Provide a user interface where integration engineers can build, organize, and test their ideas.

  • 3

    Build a solution that will help in-house divisions to explain the personalization engine's operations.

  • 4

    Involve engineers into the design process, and cocreate the solution with them to ship something witch is really tailored for them, and make them supportive in-connection with the project.

  • Research methods

    To design an interface for a machime learning engine, desinger have to understand the basic patterns, and dig deep into the domain.

  • 1

    Stakeholder interviews

  • 2

    Redifine Goals

  • 3

    User interviews

  • 4

    Build mutual dictionary

  • 5

    Persona creation

  • 6

    Journey building

  • Design methods

    To design an interface for a machime learning engine, desinger have to understand the basic patterns, and dig deep into the domain.

  • 1

    User stories

  • 2

    User flows

  • 3

    Information Architecture

  • 4

    Flow charts

  • 4

    System map

  • Wireframeing

    Some components of the user interfacce was already developed during a previous project, so we had to fit the new screen to that. Whereas it was an extreamly UI-havy project I developed a template system for the wireframeing process, with standards.

    High Fidelity to RAT and front-end development

    Partly because of the white-space solution, and the MVP concept, we decided to start develop a working code proof-of-concept, to have some hands-on-ecperience inhouse. This prototype was an empty shell, just a user interface, developed in webgl - so it wasn't able to run real data flows. But it was useful to test the riskiest assumptions of the design, on the same time we elaborate the behavioral and visual interactions of the solution.

    Next project

    Yuspify product development