Mobility impairments are the most prevalent of all disabilities, affecting the life of nearly one in 20 individuals in developed countries. In the most severe cases, they can have dramatic consequences for the social, mental and physical well-being of those impacted. Wearable robots, also known as exoskeletons, are devices that can physically help humans achieve certain tasks by compensating a disability or augmenting their capacity. Such devices have already proven to temporarily restore some physical function such as the ability to walk for individuals who had suffered a spinal cord injury (SCI). As of today, their limited performance has however prevented them from being widely accepted and used in community. Their high cost, bulkiness, weight, poor usability and limited intelligence are often regarded as a cause. Exoskeleton research and development faces several challenges which have so far limited adoption, whether in the medical field or not. These challenges include the unpredictable and complex nature of human users; the important diversity of their pathologies and needs; the long cycles inherent to hardware development and patient testing; together with the difficulty to compare devices' design and mode of action. In an attempt to cope with these challenges, this thesis introduces the lean synthesis framework. It draws on modularity and digital manufacturing to accelerate and reduce the costs of exoskeleton development process. A collection of modules is presented, along with "DualSKin fabrication", a digital fabrication method particularly adapted to the case of wearable robotics, both used by the framework. We tested the framework by developing four lower-limb exoskeletons with different use cases. TWIICE, enabling individuals with a motor-complete SCI to walk again; SPRIINT, for running assistance of individuals with transfemoral amputation; INSPIIRE, for the study of gait and postural control; WIITE, enabling individuals with SCI to perform ski-touring. TWIICE was found to be the lightest exoskeleton capable of climbing stairs and among the fastest of its category. SPRIINT was the first exoskeleton able to assist running in individuals with transfemoral amputation and the fastest exoskeleton for running ever reported. WIITE was the first reported exoskeleton allowing SCI patients to practice ski-touring. INSPIIRE provided important and translatable insights on human gait and postural control. These findings enabled the implementation of a self-balancing controller on TWIICE. The important variety of these use cases exemplifies the diversity addressable by lean synthesis. New device development and fabrication cycles as short as five weeks were possible thanks to a high degree of reusability. Development and manufacturing costs could also be reduced by leveraging on standard components sharing and recycling. The presented DualSkin fabrication method was found to reduce the production costs of prototype structural parts by 93% with respect to CNC machining while remaining competitive for small series up to 500 parts with respect to low-volume injection molding. These findings suggest that the proposed implementation of lean synthesis contributed positively to addressing all of the considered challenges without deteriorating performance. Together, these contributions advocate for lean synthesis as a new tool for both wearable robotics research and product development.