With this Future Leaders Fellowship, I will lead a research team that will build the first self-consistent data assimilation (DA) and continuous verification scheme for space weather forecasting, which is critically needed to improve our physical understanding of, and preparedness for, hazardous space weather. Space weather is a global natural hazard which can severely impact society, industry, and be a risk-to-life. It is a known risk to energy security, communications, aviation, and satellite services. Severe space weather is driven by Coronal Mass Ejections (CMEs), which are violent eruptions of magnetised plasma from the Sun's atmosphere. Cost-effective mitigation of space weather therefore relies on forecasting the arrival and properties of CMEs at Earth. Due to the potential seriousness of space weather, it is included in the UK's National Risk Register, and is planned for in the UK's Severe Space Weather Preparedness Strategy. Thus there is a crucial need to both better understand the physics of CMEs and to improve space weather forecasting capability. However, CME prediction has failed to improve in a decade of intense research, due to both knowledge gaps and observational limitations. Sophisticated computer models are used to simulate CMEs flowing through the solar wind to Earth. However, although these models are grounded in the relevant physics, they struggle to accurately represent observed CMEs. There are two key reasons for this; firstly, the starting conditions of these models are very uncertain due to observational limitations; secondly, the representation and balance of processes in the models is incorrect - indicative of our limited knowledge of physics controlling CMEs. Heliospheric Imagers (HI), such as those developed by UKRI's STFC for NASA's STEREO mission, provide the only consistent observations of CMEs and the solar wind flowing over the whole domain from the top of the solar atmosphere to Earth. These observations show CMEs being both distorted and eroded as they flow through the highly structured solar wind, but they are under-exploited in space weather research and forecasting. DA is the process of combining models and observations, accounting for the uncertainty in each, to provide a best estimate of a system's state. By assimilating a wide range of meteorological observations, DA has revolutionised the accuracy of terrestrial weather prediction, but more importantly improved physical understanding of atmospheric processes. With DARES, we will develop a HI-based DA scheme that will revolutionise our understanding of CME physics and improve space weather forecasting skill. Working at the University of Reading, my team will collaborate with colleagues at the UK Met Office Space Weather Operations Center, UKRI's Rutherford Appleton Laboratory and KU Leuven. By comparing our HI DA constrained CME simulations against observations of CMEs flowing past Earth, and state-of-the-art spacecraft observatories such as ESA's Solar Orbiter and NASA's Parker Solar Probe, we will discover the physics crucial for understanding CME evolution. In doing so, DARES will provide the critically needed knowledge and tools required improve space weather forecasting skill. DARES is timely as it will help maximise the UK's return-on-investment from Vigil, the European Space Agency (ESA) space weather monitor to be launched in 2029. DARES also directly aligns with the UK National Space Strategy to "protect and defend our national interests in" space and "lead pioneering scientific discovery" as well as Pillar 1 of the UK Severe Space Weather Preparedness Strategy.