QuantML
I built the autograd engine as well as the models.
View sourcepip install quantmllibraryThe engine and models
I built QuantML as a Python library for quantitative time series research. I wrote a custom automatic differentiation engine, then built LSTM and temporal convolutional network models on top of it. The code covers tensor operations, sequence training, and inference.
Training and backtesting
I connected the models to feature pipelines, objectives that account for the Sharpe ratio, walk forward optimization, and a vectorized backtest engine. Training and evaluation follow time order so I can examine model behavior across successive windows.
What I evaluate
I use the library to study models and evaluate alpha signals in backtests. Rolling metrics cover signal quality, risk, drawdown, turnover, and returns. I can inspect the features, training, and evaluation together when looking at a result.