CrackAlgo: An AI-Powered Adaptive Coding Platform for Personalized Learning Through Large Language Models

Abstract

The rapid expansion of programming education has created a strong need for intelligent learning environments capable of adapting to individual learner progress. This paper introduces CrackAlgo, an AI-driven adaptive coding platform that leverages the OpenAI ChatGPT API to dynamically generate personalized coding question modules, continuously evaluate user performance in real time, and construct customized learning pathways. In contrast to traditional online judges such as LeetCode and HackerRank, which rely primarily on static problem repositories, CrackAlgo continuously updates each learner’s proficiency using a dynamic skill vector. The system identifies conceptual gaps across data structures, algorithms, and dynamic programming topics and subsequently recommends targeted practice problems. The platform is implemented using a Django REST Framework backend, a Next.js and Tailwind CSS frontend, and PostgreSQL for efficient data management. A 60-day pilot study involving 120 undergraduate students demonstrated a 34.7% improvement in problem acceptance rates, an 88% increase in daily engagement, and broader topic coverage compared with static platforms. Additionally, 87% of participants reported that the NLP-based structured feedback was more helpful than conventional generic feedback. These findings indicate that CrackAlgo provides a scalable, effective, and accessible approach to personalized programming education.