About the Project

Inspiration

Choosing the right courses can often feel overwhelming, especially when balancing personal interests, career goals, credit limits, enrollment types, class timings, and course prerequisites. We wanted to build something that makes this decision-making easier and smarter for students. That’s how the idea for Smart Course Advisor came together — an AI-powered system that helps students confidently plan their academic journey.

What I Built

Smart Course Picks is a full-stack web application that:

• Collects students' interests, skills, past courses, credit requirements, and enrollment preferences.
• Matches available courses while considering prerequisites, credit limits, course timings, and course popularity.
• Generates personalized course recommendations using LLMs (Gemma:2b on Ollama).
• Visualizes the career roadmap and suggests potential job titles.
• Provides an interactive chatbot to clarify student questions about recommendations.
• Handles course seat availability dynamically.

Built with:

• Frontend: React, Axios
• Backend: Express.js, Node.js, MongoDB (Mongoose)
• AI Integration: Ollama LLMs
• Hosting: Local development with Vite, Ollama server for LLM

Challenges Faced

• Designing dynamic prompts that accurately guide the LLM to recommend courses considering multiple parameters.
• Parsing the raw LLM responses reliably into structured JSON for frontend rendering.
• Managing complex validations like prerequisite checking, time conflict avoidance, and handling full courses.
• Optimizing for faster LLM responses while ensuring quality output.
• Ensuring UI/UX was intuitive, especially for selecting past courses and finalizing recommendations.

What I Learned

• Crafting effective and structured prompts is critical for getting quality outputs from LLMs.
• How to parse and structure semi-unstructured LLM outputs into usable React components.
• Designing user-centered forms that dynamically adjust based on backend data.
• Managing real-world backend challenges like course seat availability, MongoDB operations, and error handling.
• Building a real-world, end-to-end AI-assisted project with full-stack technologies.

Built With

• express.js
• gemma:2b
• mongodb
• mongoose
• node.js
• ollama-llm-(wizardlm2-7b
• react.js
• vite