SIGIL-PS - LLM-Mediated Programming Education
Research Assistant - LLM-mediated programming education project. Developed and deployed a conversational agent using Python, TypeScript, and Docker on Microsoft Azure with ChatGPT/LLaMA APIs for 60+ student participants. Conducting laboratory studies with a 4-person research team to assess LLM-mediated learning effectiveness.
About This Project
SIGIL-PS (Stepwise Instructional Guide for Independently Learning Programming Skills) is a research project developed at NAU's RESHAPE Lab. As a Research Assistant, I contributed to this LLM-mediated programming education initiative for novice programming students.
My Contributions:
- Developed and deployed the conversational agent using Python, TypeScript, and Docker on Microsoft Azure, integrating ChatGPT/LLaMA APIs for programming education research with 60+ student participants
- Conducting laboratory studies with a 4-person research team and analyzing behavioral data to assess the effectiveness of LLM-mediated learning approaches
Backend (sigil-ps-core):
- Flask REST API for chat, feedback, personalization, personas, and users
- DSPy-based Sigil module for tutoring responses with ChatGPT/LLaMA integration
- React (Vite) chat UI served by the API or embedded in the VS Code extension
- MySQL for users, personalization, and persona data
- Docker containerization with Azure Pipelines for deployment
Frontend (sigil-ps-vscode):
- TypeScript VS Code extension implementing a chat participant for the native Chat view
- Automatic file context: current file content, selected text, and attachable files
- Personas and personalized responses based on user feedback
- Works independently of GitHub Copilot
Technical Stack:
- Backend: Python, Flask, DSPy, Docker, Azure, MySQL
- Frontend: TypeScript, VS Code Extension API, React, Vite
- AI/ML: ChatGPT/LLaMA APIs, prompt engineering
See the repository links below for full documentation and source code.
