Fracture is an important phenomena in our day to day life, from the ease of cracking open an egg after inducing a small flaw on the surface, to the mind nerving slow breaking of the charger cable. Exactly a century ago, A. A. Griffith expressed a criteria that relates the size of the critical defect in a brittle material to the critical stress required for fracture. The presence of microscopic flaws explains why the stress required to break an 'intact' material is orders of magnitude lower than the theoretical stress predicted from breaking the atomic bonds. Despite significant progress in the field of fracture mechanics since Griffith, we still do not have a complete understanding of the why, when or where the failure will happen. This is because failure is a multi-scale phenomena: An example of this is the fracture of solid wood: from molecular to cellular scale to growth ring scale and finally the breaking of a branch. Understanding of the connection between these different scales is crucial to comprehend their complex, nonlinear and spatio-temporal response. Thus going down in scale to the precursor or the structural unit within the material where the failure originates can lead to development of reliable models for better engineering of these materials. In crystalline systems, the structural precursor is a competition between motion of dislocation and grain boundary mediated plasticity. This model is not valid in amorphous systems. Furthermore, solids that have an heterogeneous structure bring yet another level of difficulty in pinpointing the precursor events. In my thesis, I studied the fracture of gels, solids that are made of two intertwined phases: a solid network and a liquid medium. In gels, the complex coupling of viscoelasticity and plasticity makes it difficult to understand the driving force behind fracture as most of the theoretical models are based on understanding of elasto-plastic systems. The importance of gels in our daily experience, from the 'mouth-feel' of yogurt to the usage of gel as biomimetic material, requires the advancement in understanding their fracture precursors. In this work we try to find fracture precursors in gel by combining shear rheology to 3D confocal microscopic observation. This means we have to use a gel system which has microstructure large enough to be observable with optical resolution. Larger microstructure implies softer materials and thus low stresses that are difficult to apply reliably. Thus the first part of this thesis consists in developing a cantilever based setup for shear rheology of very soft solids under constant stress or strain. This setup has a sensitivity of 6 mPa in stress measurement and can reliably apply strain or stress control in both shear and normal directions. The gel system we work with is sodium caseinate, a protein gel with a microstructural length scale of order 10 µm. We describe the physical chemistry, mass transport and rheological characterization of our system to develop a fast in-situ gelation technique. In the next part, we perform controlled strain and stress experiments with this system until failure. In the case of incremental control strain steps, direct microscopic visualization combined with 2D phase correlation describes the existence of a soft layer near the maximum shear zone to be the precursor to fracture. For the long duration creep experiment, the 3D stroboscopic view observes visuals of the strand breaking events which are spatially scattered. We develop analysis based on estimation of small displacement to detect these events to observe their spatio-temporal profile. We observe that strand breaking events occur way before the catastrophic failure, and we are able to detect their nucleation which leads to failure. The spatial correlation of the principle shear component leads us to discover the elastic coupling of eshelby nature i.e. the existence of a shear transformation zone as inclusion within a large elastic medium.