"The objective of this proposal is to tackle two of the greatest challenges in quantum chemistry: (i) extending the applicability of highly accurate wave function methods to large molecular systems, and (ii) developing accurate and robust multi-reference methods that can be used for studying important but very difficult problems in transition metal chemistry, catalysis, and photochemistry. Solutions to these problems have now come within reach due to three advances we recently reported: first, the steep scaling of the computational cost with molecular size can be reduced to linear by exploiting the short-range character of electron correlation (local correlation methods). Second, the accuracy, efficiency, and robustness of these local correlation methods can be strongly improved by new tensor decomposition approaches and the inclusion of terms depending explicitly on the inter-electronic distances (F12 methods). Third, the development of highly complex electronic structure theories can be greatly facilitated and accelerated by new automated tensor network evaluation techniques. We are certain that by combining and generalizing these advances the long-standing problems (i) and (ii) can be solved. We will focus especially on highly scalable algorithms in order to use massively parallel computer systems efficiently. For linear-scaling methods this means that the size of the molecules that can be treated in a fixed time will grow linearly with the number of available processors. We will furthermore explore new multi-reference ansätze and implement analytical energy gradients and response properties for local methods. Hybrid and embedding methods to account for solvent and environment effects will also be investigated. It is our priority to make our new methods as easy to use, robust, and widely applicable as possible. We believe that they will open entirely new horizons for innumerable applications in chemistry, physics, biology, and materials science."