AI-Powered Personalized IT Skill Development System

This paper presented the AI-Powered Personalized IT Skill Development System, a unified educational platform that integrates Google ADK-orchestrated LLM reasoning, adaptive text processing, real-time code evaluation, and re- inforcement learning-based difficulty adaptation within a production-grade Spring Boot and React.js web applica- tion with PostgreSQL persistence. The system simultane- ously addresses three critical decision-support gaps in edu- cation—curriculum generation, algorithmic feedback, and adaptive testing—within a single authenticated platform. The code evaluation module achieved an overall F1- score of 91.0 percent across diverse algorithmic challenges, successfully identifying logic errors and providing Big-O complexity analysis. The text processing module reliably handled transcripts exceeding 40,000 characters, generat- ing highly relevant study guides and quizzes. Crucially, the distributed microservice architecture, bolstered by Kafka message brokering, maintained sub-5-second evaluation re- sponse times under load testing conditions. The integration of a Contextual Bandit framework with Bayesian Linear Regression and Thompson Sampling pro- vided data-efficient adaptive difficulty selection without re- quiring full sequential modeling or manual hyperparame- ter tuning. The reinforcement learning component demon- strated effective adaptation to individual learner profiles with natural exploration-exploitation balance. The quasi-experimental study demonstrated a 35 percent improvement in learning outcomes, with the experimen- tal group showing higher post-test scores (84.6 vs. 68.4), fewer submission attempts (5.2 vs. 8.5), and higher course completion rates (91% vs. 72%), confirming the system’s practical viability for real-world computer science educa- tion.

1.   ”Artificial intelligence in computer programming ed- ucation: A systematic literature review,” Computers and Education: Artificial Intelligence, vol. 8, 100403, 2025. 2.   ”Enhancing personalized learning in programming education through generative artificial intelligence frameworks: A systematic literature review,” Interna- tional Journal of Advanced Trends in Computer Sci- ence and Engineering, 2025 3.   ”A systematic review of generative AI use in program- ming education,” International Journal of Information and Education Technology, 2024-2025. 4.   ”Computer science education in the age of generative AI,” Stanford Digital Learning/SCALe report, 2025. 5.   ”Intelligent Tutoring System: Experience of link- ing software engineering and programming teaching,” arXiv preprint, 2023. 6.   Scarlatos et al., ”Training LLM-based tutors to im- prove student learning outcomes in dialogues,” arXiv preprint, 2025. 7.   Feng et al., ”CourseAssist: Pedagogically appropri- ate AI tutor for computer science education,” arXiv preprint, 2024. 8.   ”Evaluating the effectiveness of LLMs in introductory computer science education: A semester-long field study,” L@S ’24 conference paper, 2024. 9.   ”Learning by Teaching: Engaging students as instruc- tors of large language models in computer science ed- ucation,” arXiv preprint, 2025. 10.   ”The effects of GitHub Copilot on computing stu- dents’ programming effectiveness and efficiency,” Re- search paper, 2025.