In-memory sensors represent a category of sensors that incorporate both data acquisition and immediate memory storage within the same physical location1. This technology enables rapid, localized data processing and retention, significantly reducing the time and energy required for data transfer between sensors and storage systems. However, merging sensory and memory units in a single device requires complex design and fabrication techniques, which can be challenging and costly.To enhance the understanding of in-memory sensing and facilitate the development of more intelligent devices, this thesis focuses on exploring material platforms whose responses to external stimuli are inherently “in-memory”. This means these materials not only detect changes in their environment but also record these changes within their structure, eliminating the need for separate memory components. Such materials can autonomously preserve sensing signals by leveraging phase change mechanisms, mechanical deformations, chemical transformations, and other transformative processes. In these scenarios, the physical or chemical alterations within the material itself serve as enduring records of the events. The first project presented in this thesis focuses on the in-memory sensing behavior of vanadium dioxide (VO2) in aqueous solutions. VO2 is a correlated-electron material noted for its phase transition from insulator to metal near 67 ºC. This study documents the non-volatile phase transition of VO2 that occurs at room temperature, driven by native electric fields at the VO2 - salt solution interface, without the need for any external voltages. The conductance of VO2 increases in a manner correlated with the salt concentration until it stabilizes, thereby enabling the in-memory sensing of the salt solution. This device exemplifies the “memsensor”, a term reported previously2 coined to describe two-terminal devices that exhibit memristive behavior and adaptability to external stimuli. We anticipate our memsensor to inspire wetware design for brain-like computing, autonomous sensing and aquatic neurorobotics with fewer memory units and lower energy consumption. As an example, the memsensor mimics the neuroplasticity in chemosensory neurons of the model organism Caenorhabditis elegans (C. elegans)3, and is used to guide a miniature boat to replicate its adaptive chemotaxis behavior in food search. The second project in this thesis broadens the scope of in-memory sensing to encompass a more diverse range of material categories. This extension is premised on the insight that any lasting alteration in a material’s properties can, at least to some extent, help decipher the associated signal through ex situ sensing methods. Unlike in situ sensing, which involves real-time data measurements directly within the environment where the data originates, ex situ sensing evaluates material responses to stimuli after the exposure has taken place. For example, the growth rings of a tree can generally indicate the relative amount of rainfall over several years. Ex situ sensing, being energy-free and operation-free, is particularly valuable in scenarios where continuous monitoring or energy inputs are impractical. However, extracting dynamic signals over time remains a challenge in this method, which this work addresses. For time-derivative signals, the approach involves approximating the signal using a Taylor series. By employing an array of materials whose response rates function as signal-dependent and are not linearly correlated, the Taylor series coefficients can be determined by measuring these materials at a specific endpoint. This concept is demonstrated through the creep deformation of three different materials to sense stress over time. For stepwise signals, such as a step function, the value of each step is obtained using a similar process. This is exemplified by the etching effects of two materials used to detect gas components ratio in plasma.