AniMagic

The interaction

AniMagic is a team prototype built around a simple interaction: a user draws an animal, and the application turns that drawing into an animated 2.5D companion in an Apple-platform experience.

My contribution

I worked on the AI and application-system parts of the prototype. My contribution covered drawing-data collection and preparation, development of the animal-classification model, Core ML integration, and the application logic that connects the model result to the drawing experience.

Connecting model to app

The model is only useful if its output fits the interaction around it. I therefore worked across the boundary between the input captured with PencilKit, the classification result, and the SwiftUI flow that presents the companion. The model was integrated through Core ML for on-device inference.

I also contributed to team discussions about how the classification result should support the ARKit/RealityKit experience. This kept the model output connected to the product interaction instead of treating the model as an isolated demo.

Prototype outcome

The result was a working team prototype that connected custom drawing input, model inference, and an Apple-platform application flow. The prototype has not been released as a production application, and this case study does not report a classifier benchmark.

What I learned

This project reinforced the importance of designing the model boundary together with the user experience. Data preparation, model output, and application state have to agree for an AI feature to feel dependable.