Google Creative LabGoogle Creative Lab 

Software Engineer at Google Creative Lab. A decade of engineering and designing AI experiences used by millions. Built alongside artists, researchers, and creative directors.

This philosophy of open-sourcing large-scale creative experiments is central to how Google Creative Lab approaches its work. Rather than letting the installation be a one-time artifact, we wanted to show how to use Chrome in a broader ecosystem of physical interactive displays.

From January to March, 2025, I worked with researchers across Google DeepMind and Google Creative Lab to produce a series of demonstrations of embodied reasoning across three robots. The project began when Fei Xia, a senior DeepMind researcher in Mountain View, built an initial prototype on top of my Gemini Live Web Console — using it to stream a robot's camera feed back through the UI as a way to see what the model was reasoning about. Once we saw his demo, our team built on top of it: more demonstrations, richer ways to surface the robot's reasoning, and — pushing past visualization — using voice as the input layer to directly drive the robot's actions.

Each robot lived in its own room at Google X in Mountain View — Atari with around six humanoids, Aloha with rows of bi-arms (cheap enough to keep many on hand) — and Omega in London. Each room had a research team built around it, but the teams flowed freely between them, lending each other hands. From Google Creative Lab, Alexander Chen was creative director and Suz Chambers led production; Samuel Lawton, Andrew Rhee and Shashwath Santosh worked closely together on directing, filming and planning the demonstrations. We shot for three days at Google X, where I moved between rooms with each robot's team, defining tool calls, writing system instructions and tuning the interface so it would read well on camera.

Infinite Wonderland uses generative AI to create a unique retelling of Lewis Carrol's classic Alice's Adventures in Wonderland. It is part of our "Labs Sessions" series for Labs.google — the successor to Experiments with Google as Google's home for experimental AI. The project uses Google's StyleDrop as a fine-tuning technique on top of DeepMind's Imagen 2 to infinitely illustrate every sentence in the book. StyleDrop can learn a new style by fine-tuning very few trainable parameters (less than 1% of total model parameters), enabling a collaborative feedback loop between artists and our team at Google Creative Lab; I repeatedly trained new fine-tuned models to help shape our characters, iterating over rounds of creative feedback.

Octet was created at Google Creative Lab using Google AI Studio and Gemini alongside Three.js, React Three Fiber, and custom GLSL shaders, and was featured at Google's inaugural Creative House at the Cannes Lions International Festival of Creativity.