FIDDLE • INTERNSHIP

Designing how designers collaborate with AI

Designing how designers collaborate with AI

Fiddle helps teams design directly in their production codebase. During my internship, I worked on the interactions that help people guide AI changes, review the results, and stay oriented as work progresses.

Fiddle helps teams design directly in their production codebase. During my internship, I worked on the interactions that help people guide AI changes, review the results, and stay oriented as work progresses.

Company

Fiddle is an AI-powered design tool that brings designers and engineers into a shared canvas connected to their production codebase. Teams can explore ideas, edit interfaces, and move from prototypes to code within one collaborative workspace.

My contribution

My work connected the core moments of the experience:

  • Starting: chat and asset navigation, searchable screen libraries, and onboarding that gave people a concrete first move.

  • Directing: frame selection, previews, state controls, and spot editing so people could point at precisely what they wanted to change.

  • Following: the turn tree and activity states that made the model’s progress visible while keeping completed work compact.

  • Collaborating: prompt queues, shared conversations, and participant cues that helped teams coordinate around one AI worker.

  • Reviewing: a docked editor for inspecting and refining components alongside the canvas.

  • Shipping: progress feedback that carried a long-running job through to a pull request.

I also explored importing existing codebases and alternative ways to summarize agent activity. These remained separate from the shipped experience.

Interested in the story behind the work?

Due to NDA and competitive confidentiality, I can’t publish internal visuals or detailed explorations here. Get in touch at vinaythorat@berkeley.edu to learn more about my contribution and design approach, including materials I’m permitted to share.

Visit Fiddle