Temperature and salinity characterize seawater properties, water mass structure, ocean circulation, and land-ocean interactions. In the East China Sea (ECS), Changjiang Diluted Water, shelf circulation, and Kuroshio intrusion shape pronounced seasonal variability, vertical stratification, and cross-shelf gradients. However, limited in situ observations and coarse reanalysis and climatological products restrict the availability of long-term, high-resolution daily three-dimensional temperature and salinity fields, hindering detailed analyses of ECS hydrographic processes. To address this gap, we develop and evaluate a reconstructed ECS temperature and salinity dataset for 1993-2020, with a horizontal resolution of 0.05 deg x 0.05 deg, daily temporal resolution, and 15 standard depth levels from the surface to 200 m. The dataset was generated using machine-learning approaches that integrate quality-controlled in situ observations, gridded products from CMEMS and NOAA, and auxiliary predictors such as Changjiang River discharge. Temperature was reconstructed using CatBoost, whereas salinity was reconstructed using HistGradientBoostingRegressor. Model performance was evaluated using a random hold-out test set and an independent 2016 validation dataset, representing the decay phase of a strong ENSO event and anomalous precipitation and runoff in the Changjiang River basin. For temperature, the model achieved an RMSE of 0.77 deg C, an MAE of 0.52 deg C, and an R2 of 0.97 in the random hold-out test set, with RMSE and R2 values of 0.91 deg C and 0.96 in the independent validation. For salinity, the model showed high predictive skill, with an RMSE of 0.36, an MAE of 0.15, and an R2 of 0.94 in the random hold-out test set, while maintaining an R2 greater than 0.93 in 2016. These results indicate that the dataset captures interannual variability rather than merely reproducing climatological patterns. Error diagnostics show that temperature errors are concentrated mainly in summer, nearshore shelf regions, and subsurface to intermediate layers, whereas salinity errors occur primarily in the near-surface Changjiang Estuary and adjacent low-salinity waters during spring and summer. The reconstructed fields represent the seasonal evolution, vertical structure, and shelf-to-offshore gradients of temperature and salinity in the ECS. They also capture key hydrographic features, including Changjiang Diluted Water spreading, nearshore low-salinity waters, and warm, saline Kuroshio-influenced waters on the outer shelf. This dataset provides a long-term, high-resolution three-dimensional temperature and salinity product for the ECS and a valuable resource for studies of ocean warming, marine heatwaves, ecosystem responses, hydrographic variability, and ocean model evaluation.