Google Partner Innovation
Google Partner Innovation is a 2-day programme focusing on the latest Google AI technologies such as TensorFlow JS and ML kit. The event was held on 29th and 30th January, a 36-hour hackathon. There were many elite companies whose problem statements were solved by developers using the Tensorflow and ML kit of Google with mentorship from Google. We built prototypes for the respective clients during this event. Clients walked away with ideas to scale them and, to portray it to the market.
Group photo with all the participants of Google PI AI for India programme.
The Core Team from Sathyabama Coding Club, Abhiram Reddy Duggempudi, Anish J, Akash M, Dewashish, Jyothirmai Thimmaraju, and TejaKummarikuntla were very happy to get invited for Google’s Partner Innovation programme at Google office in Bangalore.
The Core Team also got a chance to interact with Googlers from other countries and got various insights, to how creative engineers, co-exist to solve real world business problems. Similarly, the core team also worked to solve and got a good response from both, the client teams (Swiggy and CRED) and the partner team (Google PI).
THE CORE TEAM | Developer Student Club - SATHYABAMA
Lead - Teja Kummarikuntla | Technical Lead [ML] - Akash M | Girls Tech & Operations Lead - Jyothirmai Thimmaraju |
Technical Lead [Appdev]- Anish J | Technical & Content Lead - Abhiram Reddy Duggempudi | Community Lead- Dewasish Kumar |
Working With Swiggy:
Swiggy Gamification:
Problem statement :
We had to build a gaming experience that will be integrated into Swiggy’s application during the IPL season.
Points to remember:
● Short span of time ● Avoid Distraction
● Marketing strategy ● User Retention
With The Swiggy Team
Swiggy IPL Gamification Prototype
What we did:
● A game where user can predict the score of next 5 overs called
Swigg-ings
● He/She can bet with their in-game currency, called balls
● The winners will be gratified, which will be recommend using a ML
model for personalisation
● Bringing in people with no-cricket knowledge to play the game by
helping them predict with a ML model that will assist them
With the CRED Team
Problem Statements :
We have been provided with two problem statements :
1. Engagement : Getting users to discover and benefit from the rewards based on location personal preference and other contextual information.
2. Customer experience : Reduce the friction of on-boarding, scan user’s credit card and extract all useful information.
Working with CRED:
Along with swiggy, we also worked with CRED. To have a little brief about CRED, it’s a member-only-club company that rewards individuals for their timely credit card bill payments by providing them with exclusive offers and access to premium experiences.
What we did :
For the first problem statement:
we focused on two major paradigms of recommendation systems for user engagement, collaborative filtering and content based models. We started with creating our own synthetic datasets, one which generated latitude and longitude using random numbers (which was our location data), for a given seed point we generated 100 more.
Other, based solely on the past interactions recorded between users and items. Our overall motive was to do collaborative and content based filtering based on the radius of the points based on the user-item interactions of the second data set. We tried indulging Memory based approaches that directly works with n values of recorded interactions of content based methods, based on the available “features”.
For the second problem statement:
We’ve created a working model with a great UI to represent the visualization of the data which benefits the idea of rewards based on location and preferences. For the second problem, we made an Credit Card Scanner app which used Measurement, Vision Image label, Stats, Scanner APIs, which was completely built on Firebase using Kotlin.
We were little inaccurate in context to scanning accurate data from the credit card, (say5 out of 10 were true), yet we found out many interesting base problems that can be solved.Like, color and background of the card, font of the numbers etc.
Finally, presented it with a UI that involves details to scan instead of the traditional type system.In the end, the client was happy with our approaches and use cases.