Deploying a machine learning model usually means provisioning infrastructure, setting up APIs, dealing with authentication, cloud costs, and so on. And sometimes, all you wanted was to show someone your idea.

This talk is about skipping all of that.

We’ll walk through how to run ML models inference directly in the browser, using ONNX Runtime Web, WebAssembly, and WebGPU when needed. The result: secure, privacy-respecting, and lightning-fast deployments—served as simple static web pages.

We’ll cover:
- How to convert and export models to ONNX from your favorite frameworks such as Scikit-Learn and Tensorflow
- How to use onnxruntime-web to run models client-side with WASM or WebGPU
- A hands-on demo: turning a notebook model into a web-based ML app with no infrastructure
- How this approach accelerates iteration, demo cycles, and stakeholder feedback
- Why ONNX helps bridge experiments, demos, and production
- Extra scenarios: edge ML, offline use, and hardware-light deployments

You’ll leave with:
- A deploy-in-minutes workflow for ML demos and prototypes
- Tools to tighten the loop between modeling and feedback
- A mental model for using ONNX to unify the path from experimentation to shipping

This talk is for: ML engineers and data scientists who want to move faster, share working demos without blockers, and cut their time-to-feedback from weeks to hours.

Forget staging environments. Your next ML prototype could just be a shareable URL.