FirstGem

Inspiration

We were inspired to make FirstGem to help first-generation low-income individuals have an easier time navigating processes such as college, personal finance, and life. These individuals face many unfair barriers to success, and we believe FirstGem will help break these barriers.

What it does

FirstGem: Empowering First-Generation Low-Income Individuals for Success

FirstGem is an application that leverages technology to empower first-generation low-income individuals on their journey toward prosperity. Comprising three distinct digital tools, this project addresses financial, educational, and personal challenges, providing comprehensive support to help participants surpass obstacles and achieve their goals.

Advisory Chatbot: At the core of FirstGem is an intelligent advisory chatbot. This AI-powered companion engages users in insightful conversations about their financial, social, and academic aspirations and challenges. The chatbot offers personalized guidance on finance, academic achievement, college admissions, and more. By demystifying complex financial concepts and offering practical solutions, the chatbot equips users with the tools they need to navigate their financial landscape confidently.
Resource Repository: FirstGem features a curated repository of resources tailored to the needs of first-generation low-income individuals. This digital library encompasses many materials, including a chatbot therapist, educational guides, scholarship databases, and career development tools. Menthelp, a chatbot therapist, is another project of ours. It is a Large Language Model finetuned on therapy data that can act as a therapist. Mental health is essential for everyone, yet therapists can be expensive and hard to access. We hope Menthelp will help low-income individuals get access to therapy. Users can easily access relevant information to support their academic pursuits and professional growth.
Text Analysis Chatbot: Effective comprehension of complex texts is paramount in pursuing higher education. The text analysis chatbot allows users to input challenging texts, such as college financial aid pages, and receive simplified breakdowns. This tool helps users understand intricate information by breaking down text into digestible portions, fostering better comprehension and informed decision-making.

Together, these three technological components create a synergistic ecosystem that transforms participants' paths to success:

FirstGem envisions a future where technology bridges gaps, dismantles barriers, and paves the way for first-generation low-income individuals to thrive. By providing cutting-edge solutions to financial, educational, and personal challenges, this project propels participants toward prosperity and sets the stage for a brighter future for themselves and their communities.

How we built it

As a prototype, we imported Llama into Colab and gave it instructions on how to help the user. Then, we used the HuggingFace projects of Llama 2 and Biobert for the advice bot and analysis bot, respectively. We then built a website using Repl.it, Flask, Bootstrap, and HTML/CSS/JS.

Challenges we ran into

We ran into the challenge that the 7B version of the Llama 2 model on Hugging Face crashed, so we had to use the 13B. The 13B model performs better but is more resource-intensive and often prohibitively slow. Luckily, the 7B model returned after some time, so we used that.

We also encountered the challenge of getting the logos of the resource websites. Some were half inverted color, SVGs, etc., but luckily, Lucas knew how to work with web elements (with inspect element and more) from the user side and got all their logos into images that could be displayed on our website.

We also had the problem that we wanted the users to have examples to test on the bot, so we made buttons that allowed the user to copy a prompt or text and paste it into the bot.

Accomplishments that we're proud of

We are, first of all, proud to be able to work on this project nationally. Lucas comes from Texas, Aaron from New York, and Charan from Silicon Valley. Coordinating between different locations, schedules, and time zones was a challenge, but it came with the valuable benefit of diverse viewpoints and life experiences.

Second, we are proud to use our technical skills to work on a product that impacts underprivileged groups, such as first-generation low-income individuals.

What we learned

We learned a lot about time management since we had to coordinate from different parts of the USA, generative AI as we had to add an LLM to our project, and web development.

What's next for FirstGem

We plan to distribute FirstGem to people in our schools (all 3 of us go to different schools) who are first-generation low-income, but also as a general resource for everyone to use. We also plan to add features to make FirstGem help with other functions of career and life, like recommending people to network with on LinkedIn based on a user's needs, helping with resume work, and more.