Pod-Matcher

Making life easier at MLH, Matching mentees to their mentors!

Comment

Inspiration

Recruiters have long found manually scanning resumes to be a time-consuming activity. Once they've combed through and filtered the resumes, they'll have to pair the mentees with a suitable mentor. The inspiration for the PodMatcher comes from the hope to automate this process for MLH in particular and save time and resources.

What it does

As part of the submission to the MLH hackathon, we created an application that:

  • Asks user (presumably a Pod Mentor/MLH admin) to upload a requirements document detailing what skills/primary language/timezone they are looking for in potential mentees.

  • These requirements are then matched with a set of applications that are stored in the firebase DB

  • The ranked applications are shown in decreasing order based on how closely they fit the uploaded requirements.

  • PodMatcher is built on React therefore responsive across devices!

How we built it

This project uses:

* BoW and TF-IDF vectorizer to convert the text documents into vector form. 

* Cosine similarity as the comparison function.

* Textract python library to read word documents.

* Flask, for the NLP tasks and hosting backend server.

* Firebase Firestore for the DB

* React for the Client-side

* Heroku to host the application

* Can-Merge utility to determine whether or not a particular pull request can be merged 
without navigating to the web interface. 
The GraphQL API of GitHub is used to fetch the latest status of any status checks 
currently running and determine if the PR should be merged.

Challenges we ran into

  • We understood the instructions a day into the Hackathon due to a misinterpretation, so our ideation began on Day2.

  • Deepti had domestic Git issues, due to her PC acting up, therefore could not collaborate on GitHub

  • Another teammate, Priya unfortunately fell really sick during the Hackathon.

  • Riya meanwhile faced trouble debugging during API integration, due to pesky CORs issues.

Accomplishments that we're proud of:

  • Coordinating sync-ups due to different timezones

  • Had a tough time debugging deployment and API integration issues.

  • Despite the late start and the issues faced by the team, we pulled through and built our MVP :rocket:

What we learned:

  • Natural Language Processing methods like BoW, TF-IDF and cosine similarity

  • React-Flask integration

  • Writing a Flask backend, and exposing it as an API endpoint

  • Most importantly, the significance of communication- and drive to actualise the vision by sticking till the end:rocket:

What's next for Pod-Matcher

  • Expand the React Client functionalities to allow for the creation of multiple Pods and display in the Home Page.

  • Work with ML modelling to add weighted scores to specific parts of an application and improve matching

  • Add an admin dashboard to upload/download applications to Firestore DB

Share this project:

Updates