microsoft/qlib

Built on a modular architecture, Qlib provides end-to-end infrastructure for quantitative research including market microstructure simulation, portfolio optimization frameworks, and point-in-time databases for handling realistic trading scenarios. The platform integrates pre-built deep learning models (Transformer, LSTM variants, TabNet) alongside novel architectures like Temporal Routing Adaptor and meta-learning frameworks for domain adaptation across market regimes. It exposes standardized factor engineering pipelines and backtesting engines that interface with both historical data sources and live market feeds, enabling seamless transition from research prototyping to production trading systems.

38,683 stars. Actively maintained with 1 commit in the last 30 days.