intelliJob

Our Work Flow
our output

Inspiration

Lots of people spend way too much time looking for jobs. And it is even worse for people who require accommodations for their disabilities. There's no way to know if a job is suitable for them or not, as the vast majority of jobs do not come tagged with accessibility.

So we decided to take it upon ourselves to build the accessibility tags, even for jobs where they don't originally exist

What it does

A three step process:

The customer submits a simple form telling us what accommodations they need
Our cutting-edge NLP algorithms look up the Indeed database for recent jobs and auto-tag jobs that may be accommodating to the particular disability -We send this tagged, curated, and ranked list of jobs to the customer via e-mail and/or display it on the indeed job portal

How we built it

We built a database of jobs by scraping indeed
We manually- yes, MANUALLY, labelled a bunch of jobs
We trained two different natural language processing algorithms, with different approaches to the problem
We combined these two approaches to a single percentage result for filtering and ranking
(We had too much caffeine)
We integrated our output set of jobs and probabilities into an indeed-like front-end
(We patted ourselves on the back)

Challenges we ran into

-We had no idea of NLP before this project so it was a learning curve for all of us -Learnt it was illegal to scrape official government websites and we hope we are not on any hitlist now

Accomplishments that we're proud of

-The model predicts with high accuracy which jobs a person with disability can access/get. -Able to generate big enough and trustworthy enough dataset to predict accurately -Implemented Two-Pronged approach with Stochastic Gradient Regression and Text Clustering using Word Embedding models -Seamlessly integrates with the Indeed website

What we learned

The most significant thing we realised was how there are options like screen reader available, but no accessibility tags for people to actually understand which jobs are accessible to them
On the technological side, we learnt how to mine data from databases, understand the importance of data cleaning and labelling (making use of various processes in the NLP pipeline), and use NLP to find jobs that are accessible to people with disabilities
Web scraping with BeautifulSoup
NLP with spCy and StanfordNLP
Text Clustering using Word Embedding
(We have MAD amounts of patience)

What's next for intelliJob

Integration with the Indeed Platform to make our algorithm a part of their page. This removes the need for any web scraping.
More accessibility features on the output page, eg: colour contrasting, text reading, etc.
A good day's sleep 😇

Built With

Submitted to

iNTUition v8.0
- Winner Indeed's JobTech Track

Created by

I worked on the web scraping and database cleaning and preparation. I also worked on the pitch deck and the video.

Swapneel Bhatt
I worked on the development of rulebase by scraping online databases. Worked on integrating the rulebase with fasttext algorithm and integrate final codes

surviving-lifeform
I worked on the automated data cleaning pipeline, as well as the word embedding and clustering algorithms. I also contributed to ensuring that all our codebases integrated seamlessly with one another to ensure a quick flow of data.

Aditya Sundaresan
Jainil Sheth