Engineering Log.

2026-06-15 // ARCHITECTURE

Deprecating Redis for Custom Rust Memory Stores

We recently hit the limits of single-threaded Redis deployments in our high-frequency trading matching engine. Moving to a custom Rust-based shared memory architecture reduced p99 latency by 32%.

2026-05-02 // WEBGPU

10 Million Points at 60FPS: WebGPU OffscreenCanvas

Rendering massive financial datasets requires bypassing the main thread entirely. We open-sourced our WebGPU rendering pipeline, achieving zero-jank frame rates for enterprise institutional dashboards.