Home / Current Issue / Paper 1717030
SyntaxSensei: An Intelligent Programming Learning Platform Using Fine Tuned Language Models, Adaptive Quizzes, And Gamified Leaderboards
Subject area: Science,Engineering and Technology · Area of research: Education
DOI: https://doi.org/10.64388/IREV9I10-1717030
Abstract
The domain of software development education faces persistent challenges in effectively translating theoretical, conceptual knowledge into practical problem-solving skills, particularly for novice programmers. While the advent of large language models (LLMs) has revolutionized code generation, their application in educational settings often inadvertently bypasses the learning process, providing direct answers rather than scaffolding understanding. This paper details the architectural design, pedagogical framework, and implementation strategy for SyntaxSensei, an intelligent educational platform engineered to provide structured, self-paced programming instruction. We propose a departure from generic conversational agents in favor of a highly specialized, domain-specific tutoring framework. Our core innovation centers on the instruction fine-tuning of a high-performing, resource-efficient foundation model—specifically, the StarCoder2-3B variant—utilizing Quantized Low-rank Adaptation (QLoRA). By intrinsically linking beginner-friendly code explanations with dynamically generated quizzes and a robust gamified leaderboard module, SyntaxSensei forces active recall and validates student comprehension while sustaining high engagement levels. The system operates on a full-stack architecture prioritizing low-latency inference through FastAPI and a responsive Next.js frontend, culminating in a production-ready pedagogical tool tailored for high school learners, undergraduates, and self-taught developers.
Keywords
Large Language Models, Code Education, Parameter Efficient Fine-Tuning (PEFT), Adaptive Quizzes, Gamification, QLoRA Fine-Tuning, Full-Stack Architecture, FastAPI, Next.js, Cognitive Load Theory.
How to cite this paper
@article{1717030,
author = {Kunal Nath, Samridh Chauhan, Tanmay Jadhav, Pranit Virkar},
title = {SyntaxSensei: An Intelligent Programming Learning Platform Using Fine Tuned Language Models, Adaptive Quizzes, And Gamified Leaderboards},
journal = {Iconic Research And Engineering Journals},
year = {2026},
volume = {9},
number = {10},
pages = {3768-3774},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1717030.pdf},
abstract = {The domain of software development education faces persistent challenges in effectively translating theoretical, conceptual knowledge into practical problem-solving skills, particularly for novice programmers. While the advent of large language models (LLMs) has revolutionized code generation, their application in educational settings often inadvertently bypasses the learning process, providing direct answers rather than scaffolding understanding. This paper details the architectural design, pedagogical framework, and implementation strategy for SyntaxSensei, an intelligent educational platform engineered to provide structured, self-paced programming instruction. We propose a departure from generic conversational agents in favor of a highly specialized, domain-specific tutoring framework. Our core innovation centers on the instruction fine-tuning of a high-performing, resource-efficient foundation model—specifically, the StarCoder2-3B variant—utilizing Quantized Low-rank Adaptation (QLoRA). By intrinsically linking beginner-friendly code explanations with dynamically generated quizzes and a robust gamified leaderboard module, SyntaxSensei forces active recall and validates student comprehension while sustaining high engagement levels. The system operates on a full-stack architecture prioritizing low-latency inference through FastAPI and a responsive Next.js frontend, culminating in a production-ready pedagogical tool tailored for high school learners, undergraduates, and self-taught developers.},
keywords = {Large Language Models, Code Education, Parameter Efficient Fine-Tuning (PEFT), Adaptive Quizzes, Gamification, QLoRA Fine-Tuning, Full-Stack Architecture, FastAPI, Next.js, Cognitive Load Theory.},
month = {April},
doi = {https://doi.org/10.64388/IREV9I10-1717030}
}