Assessing students’ spoken language skills at scale remains a challenge in education. Speaking assessments typically rely on human raters, making automation a crucial step toward broader scalability. This work presents a system for automatic spoken language assessment, developed in the context of Ceibal en Inglés, a national educational program in Uruguay that provides English instruction to thousands of students in public primary and secondary schools. Designed to evaluate both speech production and semantic content in students’ responses, the system incorporates analysis components tailored to different aspects of oral performance. The speaking assessment includes exercises where students record responses to prompts, such as short questions, extended responses, and image-based tasks. For speech production, we employ a multi-task pronunciation assessment model designed to evaluate multiple aspects of spoken output. For semantic content, we developed a pipeline starting with automatic speech recognition (ASR), followed by a large language model (LLM) guided by customized prompt sets for each test part. We evaluated the system components using different datasets that reflect the linguistic and demographic profiles of the target student population. The proposed approach, which combines speech analysis and Natural Language Processing, has proven to be an effective solution for addressing the automatic evaluation of spoken language and supports ongoing efforts toward its real-world implementation in large-scale educational programs.