Finance & Trading
Quantum and classical machine learning for the problems desks have: too little data on crashes, regimes that shift and optimization under real constraints.
For quant desks and risk teams with small, regime-shifting datasets.
Design partners Sqetch Finance is a development preview. We take a small number of design partners: bring a workload and we scope it together.
Where we think it helps.
Synthetic data for stress tests
Regime-aware synthetic market data to test strategies on scenarios outside the historical window. Classical TimeGAN now; a quantum circuit Born machine noise source in the lab.
Regime detection
Classify market regimes with gradient-boosted trees, neural networks or quantum neural networks, and route trading logic per regime.
Execution and portfolios
Constrained portfolio optimization formulated as QUBO or MILP through Aria-Opt; optimal execution and interbank payment reordering are next on that path.
Walk-forward backtests
Walk-forward backtests on synthetic series, scored on PnL, Sharpe ratio and drawdown.
What you can use now.
- OHLC generationClassical TimeGAN for synthetic OHLC series, with a PCA and t-SNE report comparing real and synthetic windows.
- Regime classificationGBDT, DNN or QNN classifiers over OHLC features, behind one interface.
- HMM regime labelsLog return, range, ATR, realized volatility and momentum fed to a three-state hidden Markov model; the labels train the classifiers downstream.
- Time-ordered splitsFit on the past, score on the future. No shuffled splits that leak tomorrow's data into training.
How it would plug into your desk.
Planned for Sqetch Finance; we scope it with each design partner.
- InterfacesPython SDK, REST and Excel, plus containerized apps for backtests and QML.
- DeploymentCloud proof of concept, on-premise enclave or scale-out cluster.
- FederatedTrain across sites without moving data; only derived signals are aggregated.