Edge-AI Lab

Pick a photo. Watch your browser build a 3D depth mesh and describe the scene. Nothing is uploaded.

Edge-AI Lab

Drop a photo, snap one with your camera, or pick a sample. A depth model and a scene classifier run entirely in your tab. Nothing leaves your device.

🖼
Samples

Why on-device matters

All inference here runs in your tab — WebGPU and WebNN where your browser supports them, WebAssembly where it does not. No image ever leaves your device, there is no per-request API cost, and latency is whatever your hardware can do. The same techniques apply to kiosks, set-top boxes and mobile apps where privacy, cost or connectivity rule out the cloud.

Need ML that runs where your users are?

On-device vision and language on browsers, Android and embedded hardware.

Read the write-up