Our aim was to build a product using our diverse skillsets for people that were most affected by COVID 19. By going through 4-5 hours of research and scraping data through online resources we started performing engineering data analysis on various sectors affected by COVID 19 in terms of losing their job. The conclusion led us to believe that people performing local jobs like servers, technicians, etc. are the most affected but they don't have a medium to find new jobs, unlike others who have LinkedIn and Indeed to find new jobs.
What it does
Pandelivery is an app that is reverse engineered, where we provide local unemployed people a resource for finding jobs. The person having a job just posts a job and the person available at the nearest location with the required skillsets is assigned the job.
For example Just like how Uber helps users find a ride near them. Pandelivery will help the user having a job to be completed like delivery, cleaning the lawn, or doing some other chores can just look for available people in their respective areas.
How I built it
We used Tableau and excel to perform data analysis. Proto.io to prepare design screens. Further we just python and deep learning libraries to create a face detection algorithm.
Challenges I ran into
Our challenge was to see how can we ensure the safety norms are being practiced so that the person posting the job is not risking their health. For that, we decided to build an algorithm that uses face detection and identifies if a person is wearing a mask or not. This will ensure a person is practicing safety guidelines.
Accomplishments that I'm proud of
It was not easy to get a headstart as the topic was not specific and there were multiple ideas in our head. We got really excited once we were settled with Pandelivery. We both had skills that were transferable hence nowhere during this event were heavily relying on one particular job. It was evenly divided while making sure we both learn something new from this competition.
What I learned
For both of us scraping data from the web was something new and we had real fun doing this. Also, while working on the prototype there were some features that were had never used which was fun using to build are design screens.
While scraping the data, we could not get enough dataset for our model to predict with higher accuracy. In some time we realized we could utilize the technique of Image augmentation wherein we created more samples from the existing samples by just rotating the images giving our model more samples to train on.
What's next for Pandelivery
We would like to upgrade our mask detection algorithm by adding new features like checking sanitary conditions or if people are maintaining social distance while performing a job. Further, we would like to design and create a database system to store our data which can be used to make Pandelivery more data drive.