Inspiration
Our inspiration for this project stems from several key factors. As college seniors, we've all been feeling the heat of career anxiety, and it's only intensified with the recent layoffs and the skyrocketing cost of living. The post-graduation phase can be particularly daunting, especially with the burden of student loan debt, which averages $37,338 per borrower, according to the Education Data Initiative and a huge spike in inflation rates, peaking at 8.0% in 2022. In addition to the complexity of general life transitions, graduation can become a particularly stressful time for everyone.
More specifically, managing finances, finding the right resources, and being able to afford them are all hard things to do. Many of our friends have shared the same feeling of being lost after graduation. With these worries and issues in mind, we wanted to create a one-stop hub for new graduates— a place where they can find everything they need to kick start their post university life. As not everyone has easy access to career counselors or a supportive network of friends and family, depending on their educational background, this one-stop hub is meant to provide personalized finance, career, and mental health advice and resources given a user’s background.
A lot of the current resources out there that do something similar (Linkedin, Indeed, etc.) are not specific enough for new grads.
What it does
We've developed a comprehensive React web app designed to assist you in navigating the challenging transition from college to your professional life. Our multidimensional platform encompasses key aspects, including career planning, financial management, and mental health support. On our website, you'll find dedicated sections for Career, Finance, and Mental Health. The Career page features an interactive career recommender that evaluates your uploaded resume, presenting the top few careers that align with your qualifications and interests, complete with a confidence score to gauge your suitability.
We used Github pages, which includes setting up continuous integration in order to set up an environment to deploy our application. Our application is deployed on a .Tech domain name.
In the Finance section, we offer an interactive finance calculator that enables you to input your salary and workplace location. Our website then generates a customized breakdown to help you allocate your income optimally for short and long-term goals, factoring in the cost of living in your specific area.
Recognizing the importance of mental health during this pivotal stage, we've incorporated a dedicated resource tab. This section not only connects you to valuable resources but also delves into the mental health benefits offered by companies to provide essential support.
In addition to this website, we've harnessed Google Cloud Vision AI, utilizing a trained single-classification model to assess resume texts and recommend the most suitable career options. This model delivers high-quality results with average precision, ensuring that you're well-equipped for your professional journey. It's been trained on a diverse dataset, encompassing 10 labels (jobs) and 10 datasets (resumes and CVs) representing various industry, research, and academic experiences, requiring 4 hours and 43 minutes of training.
How we built it
To develop our project, we used a combination of technologies and tools. Our foundation was a React web app, built with Next.js and React Bootstrap, which allowed us to create a responsive and dynamic user interface. In the backend, we used AI capabilities through Google Cloud's Vision AI and Natural Language Processing AI API.
We leveraged Google Cloud's Natural Language Processing AI API for text analysis, enabling our project to offer enhanced insights and functionality. To visualize the data and insights generated by our AI-driven processes, we utilized Chart.js, a powerful data visualization library that helped us present information in an intuitive and user-friendly manner.
Challenges we ran into
During this project, we encountered several formidable challenges. First, we grappled with the limitations of our training data and the time-consuming process of labeling it. Training ten different job classifiers with only ten resumes per job proved to be an arduous task, consuming approximately five hours of our time. Moreover, resource constraints posed a significant hurdle, as we lacked the funds to access external APIs, making it even more critical for us to devise efficient, in-house solutions.
Additionally, the entire project, from ideation to website creation, machine learning incorporation, and website deployment, had to transpire within a tight timeframe. This challenge was further compounded by the fact that many of the technologies involved were entirely new to our team (which only had 2 people), adding an extra layer of complexity to the development process. Despite these obstacles, our team persevered and found innovative ways to overcome these challenges, ultimately delivering a successful project.
Accomplishments that we're proud of
Accomplishments that we're proud of include creating a personalized career recommender based on AI and ML technologies that we have learned about in class and applied here. This feature can significantly assist new graduates struggling to find their footing in the post-grad world. We also take pride in the design choices we've made for our website, including our carefully curated color palette and branding, which seamlessly reflect in our user interface which was coded from scratch. Additionally, our personalized finance tab displays nice integrations of graphs using Chartjs, and provides valuable financial insights for users.
What we learned
Training an ML model with a custom dataset came with a huge learning curve. We learned that there are very specific formats allowed for a dataset, and that it is important to have a large variety of datasets for a single label, as that strengthens the model. We also learned how to create an endpoint for the Google Cloud Natural Language Processing API, which had extensive configuration steps and we learned that it was not possible to directly add it inside our React application without a running backend. We learned in general about the development and design of components within a React application, something neither of us had much experience in beforehand.
What's next for GradHive
Moving forward, we want to continue our project these aspects. First and foremost, we plan to further enhance the website's functionality by incorporating Google Cloud Vision AI endpoint. This involves establishing a backend server and connecting it to the API endpoint we've created, allowing our React app to seamlessly access and utilize this resource.
While we've made substantial progress, it's essential to acknowledge that full-stack integration was not feasible within our current time constraints. However, this is a priority for the future.
To improve the accuracy and depth of our platform, we're committed to training Google Cloud Vision AI with more extensive data, enabling it to classify thousands of jobs with even greater precision. Simultaneously, we aim to leverage superior APIs to gather more accurate and comprehensive rent and living data, particularly for our Finance section.
Another critical step forward is increasing the document versatility when it comes to resume uploads. Currently, we can only accept text files, but we're determined to streamline the user experience by offering PDF and Word format options. Achieving this goal may involve tapping into tools like Google Cloud's Document AI, which can extract, classify, and split documents through a range of pretrained models or custom models within the Workbench suite.
Log in or sign up for Devpost to join the conversation.