Against the backdrop of China"s new-type urbanization, urban regeneration has emerged as a critical frontier for enhancing the quality of existing urban spaces and promoting high-quality development. This shift presents a profound challenge not only to urban planning and design practices but also to the traditional pedagogical models responsible for training the next generation of professionals. Conventional design education, often characterized by a blueprint-style and task-driven approach, proves increasingly inadequate in equipping students with the interdisciplinary skills and evidence-based critical thinking required to navigate the complexities of contemporary urban issues. In response to this pedagogical gap, this paper introduces and evaluates an innovative teaching and research paradigm centered on the Research by Design methodology. Using a collaborative design workshop focused on the urban periphery of Nanjing University"s Gulou Campus as a case study, this research demonstrates a novel, integrated approach to design education that systematically merges advanced analytical techniques with real-world urban practice. The core of this pedagogical innovation lies in its structured, three-stage framework that mirrors a comprehensive, cyclical design process: pre-design planning, in-design proposal development, and post-design evaluation. This framework is designed to shift the educational focus from mere outcome generation to a rigorous, reflective, and evidence-driven process. In the pre-design stage, the paradigm moves beyond conventional site analysis by integrating machine learning for a deep, human-centric understanding of the urban environment. We employed semantic segmentation on a dataset of 834 street-view images collected by students and residents to quantitatively analyze public perception and preference. This data-driven investigation revealed critical insights into the site"s socio-spatial dynamics. Key findings indicated that factors such as the functional mixture of street-level services, the comfort and accessibility of pedestrian pathways, and the abundance of landscape vegetation positively correlated with public preference. Conversely, high traffic density and imposing architectural volumes were found to have a significant negative impact. This quantitative evidence provided a robust, objective foundation for problem identification, empowering students to define design challenges based on empirical data rather than purely intuitive or experience-based assumptions. Building on this evidence-based foundation, the in-design stage guided student teams to formulate targeted urban regeneration proposals. Two primary strategic directions emerged directly from the pre-design analysis: Optimizing Pedestrian Convenience and Enhancing Functional Mix. The first proposal focused on improving the pedestrian experience through landscape interventions, widening walkways, and creating a more comfortable and coherent circulation system. The second proposal centered on activating underutilized spaces along the campus-city edge by introducing capsule spaces with diverse functions, thereby fostering greater interaction between the university, the adjacent hospital, and the surrounding urban fabric. This stage emphasized the translation of analytical findings into concrete, context-sensitive design solutions, training students in the critical skill of evidence-based design. The final post-design stage introduced a crucial feedback loop through systematic evaluation, moving beyond subjective critiques. We conducted in-depth interviews with a panel of design experts and employed Grounded Theory for a rigorous semantic analysis of the transcribed feedback. This qualitative analysis codified the experts" evaluations, revealing a consensus on several core themes: the paramount importance of pedestrian circulation and spatial comfort, the need for high-quality and well-managed green spaces, and the demand for a richer mix of commercial and recreational functions to enhance street vitality. This structured reflection process provided students with a clear, synthesized understanding of their proposals" strengths and weaknesses, internalizing expert knowledge and transforming it into an actionable basis for future design iterations. This study redefines the "Research by Design" paradigm for the contemporary urban context by holistically integrating pre-planning and post-assessment into the educational workflow. By fusing machine learning, qualitative analytics, and creative design thinking, this teaching model makes significant contributions to educational reform. It fosters a pedagogical shift from a passive, task-driven model to an active, problem-driven one; it elevates design thinking from being experience-led to evidence-led; and it transforms evaluation from a simple outcome assessment to a deep, process-oriented reflection. This approach cultivates a complete, closed-loop knowledge system of planning-design-evaluation-redesign, enhancing the scientific rigor, practicality, and human-centric focus of urban regeneration education. It ultimately provides a replicable framework for training designers capable of addressing complex urban challenges, offering a vital reference for spatial response mechanisms and decision-making in future urban regeneration initiatives.