This paper presents an algorithmic framework for educational program development in modern information systems, bridging traditional curriculum design with dynamic labor market demands. It introduces the Design Developer of Educational Programs (DDEduP), a web-based platform integrating a fine-tuned Llama 3 model with retrieval-augmented generation (RAG). The system draws real-time data from Kazakhstani job platforms (Enbek.kz, HeadHunter), the Atlas of New Professions, and professional standards, using semantic embeddings, cosine similarity, and hybrid retrieval to generate goals, extract competencies, and recommend courses. Evaluated across Computer Science, Biology, and History, it achieved F1 scores from 0.24 to 1.00, excelling in goal formulation and technical fields but facing challenges in diverse learning outcomes and interdisciplinary adaptation. The study advocates human-in-the-loop integration and discipline-specific refinements for effective AI-driven curriculum design in educational systems.