This project aims to develop next-generation computational frameworks combining machine learning (ML), and multiscale molecular simulations to enable predictive design of functional polymer materials for energy, bioelectronics, and sustainable applications. The first part of the project focuses on mixed ionic-electronic conducting polymers (MIECPs), a class of soft materials enabling simultaneous ion and electron transport, critical for applications such as bioelectronics, sensors, batteries, and energy storage. Despite extensive experimental work, fundamental understanding of key properties—such as the density of states (DOS), intrinsic capacitance, and charge transport in realistic polymer films—remains limited . We will develop a multiscale modelling framework combining atomistic and coarse-grained molecular dynamics (MD), hybrid quantum-classical (QM/MM) methods, and ML models to predict electronic structure and transport properties. A central objective is to train ML models (e.g., convolutional neural networks and graph-based descriptors) on simulation-generated datasets to enable rapid prediction of DOS and transfer integrals, thereby overcoming the computational bottlenecks of quantum-mechanical calculations. This will allow, for the first time, large-scale simulations of charge transport in realistic polymer morphologies and enable AI-driven design rules for tuning conductivity, capacitance, and functionality. The second part of the project targets lignin-based biopolymers, focusing on the generative design and self-assembly of lignin nanoparticles (LNPs). Lignin is an abundant yet underutilized renewable resource with enormous potential for sustainable materials, but its structural complexity and heterogeneity pose major challenges for modelling and design . We propose an integrated generative AI + simulation workflow, where deep generative models (RNNs and Transformers implemented in REINVENT 4) will explore the chemical space of lignin oligomers and optimize their structures. These generated candidates will be evaluated via large-scale MD simulations (GROMACS), complemented by machine-learning force fields (MLFFs) (e.g., MACE) to achieve near ab initio accuracy in describing intermolecular interactions and self-assembly processes. The simulations will reveal mechanisms of nanoparticle formation, morphology evolution, and solvent-dependent behaviour, providing predictive insights into LNP design. The project is computationally demanding, requiring large-scale GPU-accelerated MD simulations, training of deep neural networks, and integration of ML with quantum and classical simulations. Access to the Berzelius infrastructure is therefore essential. The expected outcomes include AI-driven predictive models, digital twins of polymer materials, and design principles for sustainable and high-performance polymers, contributing to the transition from trial-and-error experimentation to data-driven materials engineering.