Intel iGPT Assistant Product
Project details
-
Intel wanted to explore the usage of new AI models in their engineering pipeline. The exploration included creating a new internal tool that would utilize established models like ChatGPT to create new AI assistants that could be trained on a conversation or data obtained from a user.
-
I was brought into the project to lead a team of designers in creating components and product iterations for the iGPT platform, as well as create stakeholder presentations for the Intel team.
-
I led the design of Intel’s internal iGPT generative AI platform, creating over 100 scalable components and building a design system to streamline production. I delivered 40+ application pages that supported a staged release cycle, ensuring each phase met usability, accessibility, and performance goals.
-
I worked closely with developers to optimize the prototyping-to-production workflow, reducing friction and speeding up delivery for internal teams. Once released, the product would be integrated into the Intel workflow, reducing time spent on tasks.
The goal of this project was to create an internal iGPT tool that allowed users to have a "Marketplace" of user generated AI models. This is an example of the filter for that marketplace.
Creating a new assistant was a process that had to be thought through by a team of engineers, designers, and stakeholders. We determined what exactly was required, as well as an idea of what that would look like. It was my job to make that happen.
Of course, a dark mode needed to be created for use in all scenarios.
Once a bot was trained and ready to be saved to the marketplace, the user would be able to review the information before completing their upload.
With iGPT being an internal tool, we wanted to give the option to share the AI model with certain users in a secure manner. Some models would be trained on sensitive data, and Intel did not want all users to be able to access the data.
After creating the assistant, you would be able to browse and save assistants for easy access.
A simple to use dashboard that gave easy access to your saved assistants, active chats, and ability to search for new assistants with filters and keywords.
As part of the design process, I built out high fidelity prototypes to present to stakeholders and created active components for use in engineering demos.
Iterative examples of working through feedback and user pain points.