
Now as a computer science student, it is no wonder that you get to participate in hackathons — a LOT of them. So our college organized one in the month of September on 17th. There was a lot of tension as well as excitement as it was graded. We had to sit in Hatchery ( a place where students with entrepreneurial mindsets meet, greet and brainstorm) and code all day.
Venue: Bennett Hatchery
Date: 17th September
Time: 1pm — 1 am
Our task was to deploy any machine learning model API made using Flask on Heroku within 12 hours. Damn these buzzwords! Don’t worry I will tell you what they are.
API
API stands for Application Programming Interface. In layman terms, think of it as a waiter which enables communication between two entities (customer and chef). Here the user is the customer and chef is some other application which is offering you some functionality. Now here is a good video which can make you understand the concept better.
Flask
It is just a web framework based on Python which allows you to build web applications by providing many libraries and functionalities.
Machine Learning Model
Here I am talking about a machine learning algorithm which allows computers to “learn” and predict things of interest to us such as whether the given image is of a cat or dog, what would be the price of this house given size,area and other properties of house etc.
Hackathon
Here are some images of students brainstorming and converting their ideas into code.


It is an event where developers code at one stretch spanning over 1–2 days. The task is to develop some kind of application within the allotted span of time.
Now that we are through with the buzzwords, let us talk about what really happened that day.
My team consisting of 3 members (all of whom were my batch mates) were asked to deploy an API on OpenPose. Now what is this? It was just a model that was able to detect human body, hand, facial, and foot key points on single images. It kind of looked like this:-

Here is the link to that project in case you are interested:-
Now what followed it wasn’t as exciting. Our first evaluation was supposed to begin at around 4 pm and second and final presentation was to be given at 12 pm. Unfortunately we had already spent a lot of time organizing ourselves. We began working for real at around 2:30 pm as we had been waiting for our third team member who came late. Unfortunately we weren’t able to make it work due to the project requiring us to install some software which we weren’t able to get to work. We were just switching between different operating systems as some repositories for the given project were run on different operating systems and some commands were specific to that operating system. Now it was around 3 pm; no results so far. So we asked one of our batchmates (who was organizer) to change our project. He gave the green signal and allowed us to work on whatever project we wanted to. Sigh of relief! But it wasn’t over. So my team member looked for some machine learning projects on GitHub dealing with sentiment analysis. Oooh another buzzword! Sentiment analysis may sound something complex but it isn’t. It is the task of identifying a sentence as positive or negative (sometimes neutral as well or some other labels). So he stumbled upon one where the task was to identify whether the given statement was political, medicine related etc. He downloaded it and started working on making it work on our systems. Now what was my and other team member’s job. Well we were supposed to learn about how to deploy it on Heroku using Flask. We read a bunch of articles but it wasn’t that simple.
First Evaluation
And now it was time for our first evaluation. We went to our professors’ desk and told them how we weren’t able to get our first project to work and changed it later. He didn’t seem to mind it. But he asked us to implement more functionalities. He wanted us to use Twitter tweets as input to our model using JavaScript (another programming language used in web development). But the problem was none of us were that great at web development. But we had the whole day. So we started watching tutorials and reading articles.
4pm — 12pm
So we sat down again and started working. Now it was time for refreshments and snacks. We picked our choices of chips and biscuits and started savoring them without a care in the world. So after the break was over, we went back to work. After our team member finished working on the algorithm, we tested it and it was working perfectly but the only problem was there weren’t enough classes implemented into it. So he went back and started including more classes. Now we were done reading articles about Flask and Heroku. Now it was time to put it all to test. So we tried deploying some sample project but unfortunately weren’t able to for one reason or another. After my team member was done including the other classes, we started working on front end of our website. So with our little knowledge of HTML and CSS, we were able to develop a rudimentary front end. Now it was 8pm. SNACKS ROUND 2! We again chose our favorite refreshments and ate them (obviously). After that, we had to work on getting those Twitter tweets. So we weren’t able to get it to work the way it was asked of us but we still managed to include the Twitter “bit”. Now we were able to extract any user’s latest tweet given his/her username and predict its class. Front-end was ready, model was ready; now the only thing left was deploying it on Heroku. So we took little help from our peers and BOOM; it was done.
Final Evaluation

Now it was time for the final evaluation. We went to one of our faculty and she asked us about the algorithm implemented in the application. The algorithm we had implemented was Naive Bayes. It was an algorithm that works on probability. It considered features independent of each other and calculates the probability. So we gave a satisfactory answer to her and were done for the day.
Aftermath
So we went to our rooms and discussed with our other friends about what they did, how did their evaluation go and some other stuff. After that we called it a day and went to 3 hrs of very precious sleep.
THANKS FOR READING!
