Live link: https://flashfire.onrender.com/
FlashFire is built on a React and Next.js front end ensuring the cleanliness and reusability of code as well as allowing for easy state management of components. TypeScript ensures any compile-time type error is caught, and Tailwind CSS is chosen to simplify the total size of the full stack web application by allowing class names to be used for styling rather than full CSS files.
The backend is made of two parts:
- API layer The API layer connecting the frontend and backend is created using FastAPI, a lightweight backend framework. It was chosen over Django as this app only needed a way to bridge Next.js to the ML model without complicated setups, and it did not have a database nor cloud requiring Django's out of the box features.
- Machine learning model The machine learning model that predicts salaries based on inputs is trained using a Kaggle dataset called Data Science Job & Salary Prediction (Glassdoor). Here is the link for downloading it: https://www.kaggle.com/code/fahadrehman07/data-science-job-salary-prediction-glassdoor/notebook. We used a two-output linear regression model to output the salary as a range. Because of its small size (741 data points after processing and filtering out unknowns), there may be great variances in the prediction of salaries.