CodeBuddy: A Browser-Based Platform for Learning Data Structures and Algorithms Using AI-Driven Feedback

Data Structures and Algorithms (DSA) remains one of the most challenging subjects in undergraduate computer science education. Students frequently struggle not because the concepts are inherently impossible, but because the feedback available to them is either absent or counterproductive. Most online platforms report only whether code passed or failed without explaining why, while unrestricted AI tools tend to hand over complete answers before students have had a genuine opportunity to think. This paper presents CodeBuddy, a lightweight, browser-based DSA practice environment that uses the DeepSeek-Coder large language model to provide intelligent, structured assistance to learners as they work through problems. The platform hosts 70 carefully designed problems spanning 7 core topics, with support for Python, JavaScript, Java, C, and C++. Its central feature is an attempt-gating mechanism: students must submit at least two attempts before hints become available, and at least four before a full solution is accessible. This design draws on established learning science research and is intended to encourage genuine engagement with each problem rather than reaching for help at the first sign of difficulty. The system additionally provides real-time error analysis, time and space complexity feedback, and a six-tier progressive hint system that guides thinking without replacing it. Early observations indicate that the graduated approach nudges students toward deeper reasoning before assistance is sought.