BACK TO WORK
ACTIVE

Algorithmic Strategies

A 20-stage quantitative trading pipeline. Features are extracted, signals are evaluated for statistical significance (IC, FDR), portfolios are constructed using HRP/Kelly/MVO methods, and orders route through unified adapters across 15 exchanges.

20stages
270+features
15exchanges
3asset classes

TECH STACK

Pythonscikit-learnXGBoostYAML

The Problem

Most quant pipelines are bespoke for one asset class. I wanted something modular that could handle crypto, FX, and equities with the same core logic.

The Approach

Built a 20-stage pipeline with clear interfaces between stages. Each exchange has an adapter that normalizes its quirks into a common schema. New strategies are YAML configs, not new code.

Key Features

Regime-Aware Signals

Signals are evaluated within their applicable regime — what works in trends doesn't work in ranges.

Portfolio Construction

HRP, Kelly, and MVO portfolio methods, selectable per strategy.

Unified Exchange Adapters

One interface, 15 exchanges. Crypto, FX, equities — all routed through the same code path.

Statistical Validation

Information coefficient, false discovery rate, and traffic-light scoring on every signal.