Globally, over 90 million children with visual impairments encounter significant obstacles in accessing quality education, particularly in STEM fields where understanding complex diagrams, spatial structures, and abstract concepts is essential. Traditional resources such as Braille books and tactile graphics are often inadequate, as they fail to effectively convey intricate visual information. This disparity perpetuates learning gaps, hindering visually impaired students from achieving academic parity with their sighted peers. Addressing this issue is critical to advancing the United Nations’ Sustainable Development Goals (SDG) 4 and 10, which emphasize inclusive and equitable education. To tackle these challenges, this study developed a multimodal assistive learning system combining a 3D tactile pad and a Retrieval-Augmented Generation (RAG)-enhanced Large Language Model (LLM). The tactile pad, equipped with actuator arrays and AR-enabled depth camera calibration, transforms 2D textbook graphics into interactive 3D tactile representations. This enables visually impaired students to physically explore and better understand spatial concepts. The RAG-enhanced LLM personalizes and simplifies complex textual content, providing a tailored learning experience that improves comprehension. We introduced Visual Disability Education Assistant (VDEA) Benchmark to evaluate the performance of our system which involves two participants—one with severe visual impairment and another with total blindness. The participants completed over 90 tasks spanning mathematics, biology, chemistry, geography, and reading comprehension under experimental and control conditions. The experimental condition utilized the new system, while traditional tools such as Braille books were employed in the control condition. The results showed that the tactile pad improved spatial understanding accuracy by approximately 27%, and comprehension scores increased by 1.2 points on a 5-point Likert scale. The RAG-enhanced LLM improved text comprehension, reduced mental and physical workload as measured by NASA’s TLX evaluation, and increased user satisfaction. Qualitative feedback highlighted the system’s effectiveness in conveying abstract concepts such as molecular structures and topographic maps. This integrative platform bridges sensory and cognitive gaps in education, equipping visually impaired students with innovative tools to thrive academically. By fostering inclusivity and independence, it aligns with global efforts to create equitable educational opportunities, empowering these learners to excel in academic and professional contexts.