We have recently demonstrated the utility of AI (Transformer-Based Model) to predict the formation of multiple, new, and stable DES (n=337) from a general database of natural compounds. We have also expanded these studies to more complex models, leading to predictions of the melting point of DES with an average accuracy of 98%. Building upon these findings, participants in this project will receive training in basic machine learning techniques and in the implementation of molecular fingerprints to predict the formation as well as melting point of DES. Unlike text-based representations of chemical structures, molecular fingerprints are derived from molecular graphs, enabling calculations based on global molecular descriptors that preserve the chemical identity of functional groups. For us, the most significant advantage is that these vectors (also known as descriptors) can be used to import properties of interest, such as molecular structure and chemical features, directly from PubChem. Participants will be trained to improve our current database (containing almost 1800 examples of DES) and then train their own neural networks. Participants gain knowledge related to the architecture of the network, the structure of the databases, and how to test the model’s accuracy to avoid overfitting (where the neural net essentially memorizes the training examples but fails to generalize its ‘understanding’ of the subject matter). We expect participants to try several tweaks to optimize the algorithm’s performance, following the evolution of the loss parameter, the validation error (expected to decrease), and the training error as a function of the number of epochs. If this is insufficient, we will implement additional (more complex) strategies, as previously described. During the last three weeks of the program, students will be asked to select a target molecule (from a curated list of active pharmaceutical ingredients) and apply their algorithm to predict the components and molar ratios needed to produce at least 25 new DES. Out of those, they will be asked to make at least 10 DES integrating the selected compound and determine their stability and melting points. This exercise will allow participants not only to challenge their model but also to address a critical aspect affecting the clinical use of these DES.