1. Non-convex Tech’s automated trading system handles over 20 billion yuan in daily transaction volume, generating massive amounts of order and trade data.
2. To rapidly monitor system performance in fast-changing market conditions, we developed a statistics and analysis system using Rust and Polars - an outstanding Rust-based open-source dataframe library.
3. This article will present the key challenges addressed by the system, along with its overall architecture and implementation approach.
4. Non-convex Tech continues to innovate through open-source initiatives. Our high-performance asynchronous logging library ftlog features time-based segmentation, rate-limited writing, asynchronous I/O, and multi-destination logging by business type. We are developing RTL (Rust Trading Language), a Rust macro-based domain-specific language to accelerate the iteration cycle of strategy research, backtesting, and live trading. Additionally, we’re creating a browser-based high-performance visualization tool using Rust + WebAssembly, significantly enhancing strategy researchers' workflow efficiency.

Speakers:

Director of Non-convex AI Lab, responsible for the development and iteration of large-scale distributed machine learning strategies in search, recommendation, and other domains. Graduated from Peking