Multi-task learning can improve indoor location estimation using Wi-Fi signals. The method uses a single model to jointly predict latitude and longitude from received signal strength indication (RSSI) measurements. This provides better accuracy than separate models for each dimension. The model accounts for non-line-of-sight signals and augments the training data. Evaluation shows the multi-task model achieves the lowest mean error compared to other techniques like XGBoost and random forests. Future work could exploit line-of-sight information and spatial/temporal dependencies to further enhance location accuracy.