Market data
Live equity data and WebSocket price streaming feed the research workspace.
FEATURED PROJECT / QUANTITATIVE SOFTWARE
AEQUITAS.
A research workspace for turning market data into quantitative insight.
THE IDEA
Connect the research workflow.
Go deeper on the computation.
AEQUITAS brings market-data ingestion, financial models and AI-assisted strategy exploration together. My C++20/pybind11 work focuses on accelerating quantitative feature engineering within the Python-based platform.
Financial models
Risk and pricing methods including Monte Carlo VaR/CVaR, Black–Scholes Greeks and Kalman-filtered pairs trading.
Research workflows
A LangGraph pipeline supports strategy generation and critique alongside a Next.js interface.
C++20 / PYBIND11
Accelerating feature engineering.
The systems work behind my CppCon 2026 poster.
CPPCON 2026 POSTER
Accelerating Quantitative Feature Engineering — A C++20/pybind11 Extension for a Python-Based Trading Research Platform
The presentation explores moving quantitative feature computation into a C++20 extension while retaining a Python research workflow through pybind11.
This work is part of AEQUITAS, connecting my interest in quantitative software with C++ systems engineering.
INTEGRATION AT A GLANCE
- PythonResearch & orchestration
- pybind11Language bindings
- C++20Quantitative feature computation
Conceptual integration, not a benchmark.
TRY A SMALL IDEA
From prices to features.
Move the window. Watch the rolling mean respond.
INTERACTIVE EXPLAINER
One feature, step by step.
A rolling mean averages a fixed window of recent prices. A larger window smooths more of the variation, while responding more slowly to changes.
2 samples 16 samples
Synthetic data, calculated in your browser. This illustrates a feature-engineering concept; it is not the AEQUITAS C++ extension or a performance benchmark.
Price Rolling mean
- Latest price
- 110.40
- Latest mean
- Complete windows
View the sample data
100.00, 100.80, 100.30, 101.50, 102.10, 101.70, 102.40, 103.60, 103.10, 102.50, 103.40, 104.20, 103.80, 105.10, 104.70, 104.00, 103.50, 104.60, 105.80, 105.20, 106.40, 105.90, 107.10, 106.60, 107.80, 108.10, 107.40, 108.60, 109.20, 108.70, 109.80, 110.40
The first window − 1 samples have no complete rolling mean.
ENGINEERING NOTES
The interesting part is in the details.
The extension creates a useful setting for discussing the boundary between Python productivity and native computation: data representation, interface design, numerical correctness and the cost of moving data between runtimes.
The CppCon poster documents the kernel algorithms, zero-copy NumPy integration, numerical checks and measured workloads. Read it above for the implementation and benchmark context.