Pyrite

Inspiration

Fraudulent job postings can be used to scam people and harvest information. With unemployment rising, job hunters are especially desperate for jobs, and too-good-to-be-true offers can be tempting. I have even received emails from my school's career service advertising a fraudulent opportunity. If we could predict based on the information in a job posting whether it is real or fake, we could protect people's privacy and ease their minds in this difficult job market.

What it does

Pyrite allows a user to input the information contained in a job posting, then analyzes the text and warns the user if it detects fraud.

How I built it

I used this dataset from Kaggle: https://www.kaggle.com/shivamb/real-or-fake-fake-jobposting-prediction which contains metadata on real and fraudulent job postings. I used the EazyML API in Jupyter Notebook to train models to predict which class a posting falls in. For the app, I used React locally to create a form, with routers to navigate between the home page and the functionality.

Challenges I ran into

I had some challenges using the API with this large, mostly text-based dataset, and with integrating the machine learning component into the app. This was my first app, and I initially tried to make it in Python using Kivy, then pivoted and made a React app. This meant I ran out of time to connect those two aspects of my project.

Accomplishments that I'm proud of

I'm proud to have learned React and built my first app in 36 hours! This is also my first time downloading and using a Python API for an external service, so I'm proud that I stepped out of my comfort zone and learned so much from HackTable.

What I learned

I learned a lot about the process of app creation and about machine learning for business. I got a taste of how much an AI app developer needs to consider, especially in terms of user experience, and I think I finally understand what an API is!

What's next for Pyrite

The next step will be integrating the EazyML model into the app so that real users can analyze job postings! I will need to find a way to translate the limited data from the users into something the model can predict from. I would also consider adding a database and keeping track of which jobs are in fact fraudulent so that the model can keep learning, and use the dataset to find predictors of fraudulent postings.

Built With

• css
• eazyml
• html
• javascript
• jupyter-notebook
• python
• react