The Browser Can Compute. You Just Never Asked It To.

We don’t usually associate the browser with real computation. No one expects it to solve systems of equations, perform linear algebra with BLAS or LAPACK, or handle large-scale numerical operations — and yet, that’s exactly what it can do now. Using only JavaScript.

In this talk, I’ll show how we’re building a scientific computing ecosystem that runs directly in the browser — no Python, no servers, no installs. From ndarrays and ufuncs to statistical modeling and vectorized operations, we’re reconstructing the building blocks of modern computation on the world’s most accessible platform: the web.

This will be a demo-heavy talk. We'll walk through live use cases of scientific computing done entirely in the browser, including:

• Training lightweight AI models in-browser

• Running OCR using WebAssembly and JavaScript

• Toxicity detection using in-browser NLP

• Accessibility applications powered by local inference

• Porting entire game engines to WebAssembly

Talk Flow:

1. Why compute in web browsers?

   • Accessibility: no installs, no configs, just a browser!

   • Privacy and security: data never leaves your device.

   • And some more factors.

2. Live Demos, as described above.

3. The Current State of Web-based Scientific Computing

   • Brief history and current landscape

   • Limitations of existing tools and how far we've come

4. What is Needed to Build a Scientific Computing Ecosystem on the Web?

   • Core primitives like ndarrays, broadcasting, and linear algebra

   • Statistical routines, distributions, and file utilities

   • Visualization and interactivity

   • Performance via SIMD, WebAssembly, and multithreading

5. How stdlib.js Meets These Needs

   • A modular, high-performance standard library for numerical computing in JavaScript

   • Covers statistics, linear algebra, distributions, complex numbers, and more

   • Designed for the browser and Node.js

6. Features That Make stdlib.js Powerful

   • Typed array support and n-dimensional data structures

   • BLAS-like operations in pure JavaScript

   • Extensive mathematical and statistical routines

7. WebAssembly Integration

   • When and why to use WebAssembly alongside JavaScript

   • Examples involving OCR, linear algebra, and inference engines

   • Discussion of performance gains from C-compiled modules

8. Benchmarks and Comparisons

   • Performance comparisons: stdlib.js vs math.js vs Pyodide

   • JavaScript vs C vs WebAssembly for numerical tasks

   • Plots and insights on performance trade-offs and optimizations

This talk encourages developers to rethink the browser’s role in computation-heavy tasks. By shifting scientific computing to the browser:

• We unlock large-scale education without installation barriers

• We make computation accessible on any device

• We enable private, local-first, and responsive apps

• We highlight the role of open source in democratizing scientific computing

• You can run BLAS-powered linear algebra in your browser tab, using only JavaScript.

• We ported scientific code from C to JS… and it runs without a server, compiler, or dependency.

• Your browser can classify digits, analyze text, and process images, locally, offline, instantly.

• This isn't WebAssembly. This isn't Python-in-the-browser. This is raw JavaScript doing serious computation.

• The web isn’t just for rendering anymore — it's becoming a full-blown compute runtime. You’ve just never used it that way.

• You can even perform Fast Fourier Transforms, using JS, inside your web browsers!