Working memory is a central cognitive ability crucial for intelligent decision-making. Recent experimental and computational work studying working memory has primarily been carried out using categorical stimuli, rather than ecologically-relevant, multidimensional naturalistic ones. Moreover, studies have primarily investigated working memory during single or few number of cognitive tasks. As a result, an understanding of how naturalistic object information is maintained in working memory in neural networks is still largely lacking. To bridge this gap, we developed sensory-cognitive models, comprising of a convolutional neural network (CNN) coupled with a recurrent neural network (RNN), and trained them on nine distinct N-back tasks using naturalistic stimuli. By examining the RNN’s latent space, we found that: 1) Multi-task RNNs represent both task-relevant and irrelevant information simultaneously while performing tasks; 2) While the latent subspaces used to maintain specific object properties in vanilla RNNs are largely shared across tasks, they are highly task-specific in gated RNNs such as GRU and LSTM; 3) RNNs embed objects in new representational spaces in which individual object feature axes are orthogonalized compared to the perceptual space, enhancing separation of features; 4) Across time, the model preserves the geometric structure of the latent subspace representing various object properties. Interestingly, the rotational dynamics governing the transformation of WM encodings (i.e., embedding of visual inputs in the RNN latent space) into memories was shared across stimuli, yet the transformations governing the retention of a memory in the face of incoming distractor stimuli were distinct across time. Our findings indicate that goal-driven RNNs employ orthogonalized \& chronological memory subspaces to track information over short time spans. This provides a basis for formulating testable predictions regarding the processing of natural object information in the cortex during working memory.