This is our landing page where the tagline/hero, sets the tone immediately.
This is how our AI asks questions to know the person more in detail
This is the career suggestion card where the AI suggests and provides a description of a career.
This is the Paths page where you can look at the saved career in detail.
This is a reflective my progress page in which the AI shows the key qualities it saw in the person.

PathFinder

Created by: Kiara Sachdeva, Kavya Pandya, and Saumya Dugala

Inspiration

With so many new jobs opening up over the last decade, it has become quite difficult for high school students to accurately choose a career path that best fits their goals, interests, and skills. As a result, students are now noticeably more overwhelmed as they transition into work, as many of them enter fields that don't align with their personal values or abilities. Unfortunately, our team has also struggled with this issue for quite a while, and we wanted a solution. This is why we created Pathfinder, an AI website designed to help students discover their perfect career path after high school.

What It Does

Quick Overview: Pathfinder is an AI-powered website that helps high school students discover career paths that align best with their values, strengths, and interests.

How It Works: A student first logs onto Pathfinder. Then, they participate in a conversation with the AI, similar to a survey. Pathfinder asks small, specific questions about the student's interests, financial goals, and ideal work environment. At the end of the conversation, Pathfinder uses back-and-forth logic to recommend career paths that fit them, while factoring in the current job market to support the student's long-term success.

In addition, the website states why each career matches the student, helping them understand their ideal pathway. A student can also bookmark careers and access them later on the Saved Paths page to view detailed information, including:

Required education
Necessary skills
Salary expectations
Programs to pursue

Throughout the conversation, the AI is built to maintain a supportive, encouraging tone so students feel comfortable being vulnerable about their uncertainty. Our team also developed a built-in Progress Page that highlights patterns identified by Pathfinder during the conversation, helping ensure the recommendations stay accurate and grounded in what the student actually said.

How We Built It

We built PathFinder using a combination of front-end and back-end technologies.

Front-End

HTML for the structure of the website
CSS for styling and design
JavaScript for interactivity and communication with the AI
Firebase Client SDK for sign-in, kept consistent with the backend's auth verification

The front-end is what users see and interact with while using the website.

Back-End

Python for handling the AI logic and application functionality
Flask as the web framework
SQLite via Flask-SQLAlchemy for the database
Firebase Admin SDK for authorization/token verification
python-dotenv for environment variables
Groq AI API to generate intelligent, personalized responses
Environment variables to securely store API keys and sensitive information

Platform / Infrastructure

Local development via a Python virtual environment (venv)
No cloud deployment configured yet — currently a local-only setup

Development Tools

Visual Studio Code for coding and development
GitHub for version control and team collaboration

The front-end and back-end were connected so that user responses could be processed by the AI and returned as personalized career recommendations.

Challenges We Ran Into

Some of our biggest technical challenges came from the backend.

Early on, folding the system prompt directly into the same messages list that got returned to and resent by the frontend caused it to duplicate every time the conversation grew. We fixed this by separating a clean transcript — the messages that are stored and returned — from a request-only payload, full_messages, which is built fresh on each call by prepending the system prompt. This kept the system prompt out of what gets persisted and sent back to the client.

A related challenge was figuring out when the AI should stop asking discovery questions and start suggesting careers. Our first attempt used a string check, requiring replies to start with "READY:" as a signal. That marker ended up leaking into what the user saw. Rather than patching around it, we redesigned the entire response contract around structured JSON output — each reply is either a "question" or a "suggestion" object — which made the signal explicit and eliminated the leaky marker text entirely.

We also struggled with:

Connecting the front-end and back-end components of the project
Learning how to work with APIs and securely manage API keys using environment variables
Designing prompts that allowed the AI to ask meaningful questions and provide useful, accurate career recommendations
Ensuring smooth communication between the website and the AI model, which required extensive testing and debugging

Accomplishments That We're Proud Of

Successfully creating a fully functional AI-powered career guidance website
Integrating the Groq AI API into our application
Connecting the front-end and back-end so users could receive personalized recommendations
Learning how to collaborate effectively using GitHub
Creating a clean, user-friendly interface that helps students easily navigate and find careers they're genuinely interested in and capable of pursuing

What We Learned

Through this project, we learned:

How APIs work and how to integrate them into a web application
How front-end and back-end systems communicate
The importance of environment variables for security
How to use GitHub for collaborative software development
How AI can be used to solve real-world problems and provide personalized user experiences

What's Next for PathFinder

Add more detailed career information, including salary ranges, education requirements, and job outlooks
Create personalized learning roadmaps for recommended careers
Allow users to save their results and track their progress
Improve the AI's accuracy by asking more detailed, targeted questions
Add support for colleges, universities, apprenticeships, and certification programs related to recommended careers
Develop a more advanced and polished user interface

Built With

python html css javascript groq-ai-api visual-studio-code github environment-variables

We chose to submit our project to Track 4, as our solution focuses on helping students make informed decisions about their future through technology and AI-driven guidance.

Contributions

Kiara Sachdeva — Team Leader, Back-End Developer Led the team and coordinated the overall project direction and timeline. Built the Flask backend powering Pathfinder's AI logic, including the /career-advice, /career-details, /progress, and /save-career endpoints, and designed the structured JSON response contract that signals when the AI should ask questions versus suggest careers. Also led the Firebase integration, setting up Google sign-in and email magic links, and implemented backend token verification to securely identify users.

Kavya Pandya — Front-End Developer Designed and built the entire front-end experience, including the landing page, chat interface, career suggestion cards, saved paths view, and progress page. Handled connecting the front-end to the backend API and managing conversation state across the chat flow. Focused on making the UI feel warm and approachable rather than clinical, given the emotional nature of the topic.

Saumya Dugala — Writer, Tester, Supporter Tested the conversation flow end-to-end to ensure the AI's questions felt natural and its recommendations made sense, catching bugs and usability issues along the way. Wrote the project documentation and Devpost submission, translating the technical work into a clear, compelling story. Supported the team throughout development with feedback on tone, usability, and overall direction.