Exploring hybrid quantum-classical approaches for algorithmic trading
By Quasareum — Quantitative Finance × Quantum Computing × Reinforcement Learning
Quasareum is building next-generation automated trading systems that combine classical machine learning, reinforcement learning, and quantum computing to exploit market microstructure inefficiencies on decentralized perpetual exchanges.
This repository documents our research journey from quantum computing fundamentals to production-ready hybrid quantum-classical models applied to real-world crypto derivatives trading.
Implementations and experiments using Qiskit (IBM), Cirq (Google), and Q# (Microsoft Azure Quantum) — covering quantum circuits, gates, entanglement, and core algorithms.
Variational Quantum Circuits (VQC), quantum kernel methods, and quantum transfer learning using PennyLane — benchmarked against classical baselines on financial data.
Hybrid quantum-classical classifiers for identifying market regimes (trending, mean-reverting, high-volatility) on ETH/BTC perpetual futures using orderflow and microstructure features.
Quantum kernel methods for detecting non-linear cross-asset relationships and exploitable spread deviations across correlated crypto pairs.
VQC-based policy networks as function approximators in RL agents for adaptive market making and dynamic portfolio allocation.
quantum-finance-research/
├── 01-qiskit-foundations/ # IBM Qiskit circuits, algorithms, finance modules
├── 02-cirq-experiments/ # Google Cirq basics, QAOA, optimization
├── 03-pennylane-qml/ # VQC classifiers, quantum kernels, transfer learning
├── 04-crypto-regime-detection/ # Hybrid QML vs classical regime classifiers
├── 05-quantum-stat-arb/ # Quantum kernels for cross-asset arbitrage
├── 06-quantum-rl-trading/ # QRL agents for market making
├── notebooks/ # Standalone Jupyter notebooks & explorations
├── utils/ # Shared utilities, data loaders, feature engineering
├── data/ # Sample datasets (raw data excluded via .gitignore)
└── requirements.txt # Python dependencies
| Layer | Tools |
|---|---|
| Quantum Frameworks | Qiskit, PennyLane, Cirq, Q# |
| Classical ML/RL | PyTorch, LightGBM, Stable-Baselines3 |
| Data | Hyperliquid WebSocket, Binance API, TimescaleDB |
| Execution | Hyperliquid Python SDK |
| Infrastructure | Python 3.11+, Jupyter, Docker |
| Credential | Institution |
|---|---|
| CS50 AI | Harvard University |
| Financial Markets | Yale University |
| Quantitative Finance | NYIF |
| Azure Quantum Computing | Microsoft |
| Quantum Business Foundations | IBM (Credly) |
| Quantum Information Fundamentals | IBM (Credly) |
| Quantum Algorithms | IBM (Credly) |
| Quantum Machine Learning | IBM (Credly) |
| Investment Foundations | CFA Institute |
We follow a rigorous experimental methodology:
- Hypothesis — Define what quantum advantage we expect and why
- Classical Baseline — Build the best classical model first (LightGBM, neural net)
- Quantum Implementation — Build the hybrid quantum-classical equivalent
- Fair Comparison — Same data, same walk-forward validation, same metrics
- Decision — Integrate quantum only if it demonstrably outperforms
No quantum hype — only quantum results.
All experiments are evaluated on:
- Sharpe Ratio (risk-adjusted return)
- Maximum Drawdown (worst-case loss)
- Profit Factor (gross profit / gross loss)
- Stability (performance consistency across validation windows)
| Phase | Status |
|---|---|
| Quantum foundations (Qiskit, Cirq, Q#) | 🔄 In progress |
| QML experiments (PennyLane) | ⏳ Upcoming |
| Crypto regime detection | ⏳ Upcoming |
| Quantum stat arb | ⏳ Planned |
| QRL market maker | ⏳ Planned |
| Hyperliquid vault deployment | ⏳ Planned |
MIT License — See LICENSE for details.
- Company: Quasareum (SASU) — Paris, France
- Focus: Algorithmic Trading × Quantum Computing × DeFi
- Platform: Hyperliquid
"The goal is not to predict the future — it's to find statistical edges invisible to the human eye and exploit them at scale."