Software engineer building at the edges of the modern web — WebGL rendering, WebAssembly performance, and AI-native workflows. Currently focused on agentic systems and the tooling that makes them feel inevitable.
Volume 01
High-performance data primitives for browser visualization — Rust/WASM downsampling, binning, spatial indexing, and streaming aggregation, plus an MCP server for AI agents. It draws nothing. Feed it a million points, get back what matters.
Open source · MIT
npm install @vizcrush/core
A headless Web Components design system built for flexibility and framework-agnostic integration. Provides unstyled, accessible primitives that work anywhere.
In active development
API development toolkit with mock server, request recording, replay, and browser-level interception. Record real API traffic, replay it offline, test without backends.
In active development
Rust's +simd128 produced 205 static SIMD instructions in my aggregate module, with none in the Welford loop I measured.
vizcrush launched with its three loudest performance claims deleted. What measuring WASM SIMD, cross-engine WASM, and WebGPU compute actually showed.
I argued that the structure of an agent run beats its length. Then I built a tool to measure it, and won't ship it until the central claim is proven. Here's the build, and the gap.
Start a project
pallav@debugdiary.dev