C++ SYSTEMS · CPPCON 2026
Python research. C++ performance.
AEQUITAS · Poster Presentation
Presented AEQUITAS at CppCon 2026: accelerating quantitative feature engineering with a C++20/pybind11 extension. The poster covers seven kernels computing 19 features, zero-copy NumPy buffers, explicit GIL release and numerical checks against pandas.
Accelerating Quantitative Feature Engineering — A C++20/pybind11 Extension for a Python-Based Trading Research Platform




