The ViEWS project started up in January 2017 with funding from the European Research Council's Advanced Grant scheme. ViEWS will develop, test, and iteratively improve a early-warning system based on the statistical analysis of data that is rigorous, transparent, and publicly available to researchers and the international community. It will provide early warnings for the three forms of political violence recorded by the Uppsala Conflict Data Project: armed conflict involving states and rebel groups, armed conflict between non-state actors, and violence against civilians. It will assess the risks of forced population displacement, and apply all of these outcome variables to specific actors, sub-national geographical units, and countries. The project will base the forecasting system on a number of theoretically informed models of factors that contribute to such violence, and of the dynamics of violence-related events, as well as models social interaction more generally. The ambition is to provide updated forecasts monthly. To achieve these ends, the project is in the process of setting up a large database that holds all the input data as well as stores the forecasts and files needed to transparently documenting them. The geographical units for which it has data divides the world into about 70,000 cells, and the system stores/processes a sizeable number of indicators for each of these for every month over a 30-year period. The other units of analysis are less numerous, but inter-level linkages adds other layers of complexity to the database. One computationally intensive task of the project is to train a number of statistical models that allow generating probability distributions for the outcomes based on the predictors in the database and projections for exogenous variables. Another intensive task is to run a program that produces n-step forward forecasts of conflict escalation based on a subset of these models, employing what can be summarized as `dynamic' Monte Carlo simulation. A third task is to integrate the forecasts from the individual models by means of Ensemble Model Averaging. The combination of large datasets and computationally intensive methods requires the use of SNIC resources.