KnightSource

Agentic AI-Driven Financial Advisor for UCF Students

Comment

Inspiration

The average student can graduate without using any of the available resources but the library and gym. KnightSource solves this problem by educating students on what they are leaving on the table and how to make the most of their college experience.

What it does

KnightSource democratizes information on resources that financially benefit UCF students. Our project helps students utilize resources that are not well advertised, by enabling them to easily have access to lawyers, healthcare options, scholarships, conferences, traveling to competitions and more.

How we built it

KnightSource is a web application that uses advanced AI features to drive the awareness of opportunities and financial literacy. The front end was built with React, Typescript, Next.js, and TailwindCSS. The backend was comprised of a Python web server built with fastapi. Our AI features included RAG through VertexAI, Google's Embedding Models, Gemini API's structured output, Gemini Computer Use, and Google ADK.

Challenges we ran into

All four of us are AI/ML specialist engineers, so learning the importance of building a pretty, user-intuitive frontend was crucial.

Accomplishments that we're proud of

We are proud of using many state of the art models and tools that we have never worked with before, such as VertexAI, Retrieval Augmented Generation, and Gemini Computer Use. While also working on an Agentic AI solution which we have worked with before but now have strengthened our skills in.

We are also proud of our React based web application because we are all backend enthusiasts and enjoyed it. But what we are most proud of is our persistency in delivering a fully working project after many attempts over previous hackathons.

What we learned

We learned so much about front end development, especially the modularity of React projects. We got the opportunity to work with state of the art AI technologies that we have not worked with before, especially Google Computer Use which was released extremely recently. We worked with a vectorized database built by embedding models on Google's VertexAI.

Most importantly we learned how to think from a user-first approach and communicate with one another throughout the entire competition. I am very proud of my teammates and love them so much!

What's next for KnightSource

KnightSource has the potential to truly be scaled to a professional application if it was connected with a functional UCF database. We would ideally like to work to build an Admin mode so that UCF administration can see what categories of resources are being utilized the most and which could be better allocated elsewhere. Finally we would also like to add more Agentic AI features utilizing Google ADK's parallelized agents.

Built With

Share this project:

Updates