AI in the Browser

Building Privacy-Preserving Local Intelligence with WebGPU, WebAssembly, and On-Device LLMs

1. About this book

Introduction: The Browser as an AI Platform

1. Why This Matters
2. What You Will Build
3. Prerequisites and Approach
4. How to Use This Book

Chapter 1: The Browser as an AI Platform

1. Why Local AI Matters
2. What the Browser Can Actually Do
3. The Three Pillars
4. A Working Example
5. What This Book Will Build
6. Summary

Chapter 2: Browser Architecture and AI Workloads

1. The Browser Engine
2. Memory Management in the Browser
3. Concurrency and Web Workers
4. Storage for Large Models
5. Browser Security Model
6. Summary

Chapter 3: WebGPU Fundamentals

1. Why WebGPU
2. Device Initialization and Adapter Selection
3. Buffers and Memory Layout
4. Compute Pipelines
5. Building a GPU-Accelerated Matrix Multiplication
6. Summary

Chapter 4: Shader Programming for Neural Networks

1. The Compute Shader Model
2. WGSL Basics
3. Vector Operations in Shaders
4. Matrix Multiplication Shader
5. Tensor Operations for Transformers
6. Summary

Chapter 5: WebAssembly Fundamentals and Toolchains

1. What WebAssembly Is and Is Not
2. Compiling C/C++ with Emscripten
3. Rust for WebAssembly
4. Wasm Memory and Typed Arrays
5. A Numerical Kernel in Wasm
6. Summary

Chapter 6: Interoperability : JavaScript, Wasm and WebGPU Together

1. Data Flow Between Layers
2. Shared Buffers and Zero-Copy Strategies
3. Calling Conventions
4. Architecture Patterns
5. Building a Tensor Library Skeleton
6. Summary

Chapter 7: Transformer Architecture for Inference Engineers

1. The Forward Pass
2. Self-Attention Mechanics
3. Key-Value Caching
4. Layer Normalization Variants
5. Positional Encodings
6. Summary

Chapter 8: Model Formats, Quantization and Loading

1. GGUF Format
2. Quantization Fundamentals
3. Loading Models in the Browser
4. Model Caching Strategies
5. Choosing a Model for the Browser
6. Summary

Chapter 9: Building a Local LLM Inference Engine

1. Project Setup
2. Tokenizer Implementation
3. Model Loading and GPU Buffer Management
4. WebGPU Compute Kernels
5. The Inference Loop
6. Sampling Strategies
7. Main Application Bootstrap
8. Expected Behavior and Running the Project
9. Summary

Chapter 10: Streaming Generation and User Experience

1. Streaming Tokens to the UI
2. Web Workers for Inference
3. Conversation State Management
4. Abort and Control Flow
5. Building a Chat Interface
6. Summary

Chapter 11: Performance Optimization and Debugging

1. Profiling Browser AI
2. Memory Optimization
3. GPU Optimization
4. Latency vs Throughput
5. Cross-Browser and Hardware Compatibility
6. Summary

Chapter 12: Production Deployment and Future Directions

1. Offline AI PWAs
2. Progressive Enhancement
3. Privacy and Security
4. Packaging and Distribution
5. The Future of Browser AI
6. Summary

Conclusion: The Local Intelligence Paradigm

References