Inspiration

Both me and my partner are high school students. As high school students, we always hear about opening a club, getting an internship, joining robotics, getting good grades, winning competitions, and more and more. There seems to be thousands of things to do in 4 years, and a college counselor helps a lot with this. However, college counselors are extremely expensive. This caused us to develop this free AI college counselor.

What it does

This program takes in a student's courses, extracurriculars, grade, and dream college, generating a roadmap based on it. The user can also upload a resume. After that, the user can chat with the AI to understand and modify the roadmap to better suit them.

How we built it

We built the program centered on Gemini 2.5-Flash as our LLM. We also used Google Search Grounding to help keep the AI up to date. On top of this, we used Gradio to build the interface, implementing an upload button and generate road map button for users to build the roadmap when truly ready. We also added a chat area for users to consult the road map and create a better tailored plan. We also used ChatGPT to help code in certain areas.

Challenges we ran into

Some challenges we ran into was building the UI, as neither was experienced with UI building. However, with the help of ChatGPT, we learned how to build a proper UI. Another issue was the upload button, which was very confusing to work with. However, after carefully analyzing the code, we realized our errors.

Accomplishments that we're proud of

Our biggest features were:

  • Our AI guardrail
  • The fact our chatbot has memory retention
    • It remembers user's data and adjusts the roadmap accordingly
  • The fact that users can upload PDFs and make it compatible with text
    • Many people have resumes they created for other programs, allowing the user to have a convenient option instead of typing it up
  • The Google Search Grounding feature
    • It keeps the LLM's information up to date.

What we learned

Development using AI still can have difficulties. We tried using Google AI studio, Gemini, and Cursor to create a UI before finally getting a proper response with ChatGPT. We also learned the intricacies of prompt engineering, as our guardrail needed multiple instructions for it to work.

What's next for AI Counselors

AI counselors would benefit heavily from a streamlined database or easily accessible source of information to access everything for college counseling much quicker. Additionally, improvements in AI models would help speed up the outputs for our project, since generating roadmaps took a noticeably long time of around 1 minute.

Built With

Share this project:

Updates