The Fifth Elephant 2019

Gathering of 1000+ practitioners from the data ecosystem

Tickets

Feed Generation @ShareChat

Submitted by Ayush mittal (@ayushmi) on Monday, 24 June 2019

Session type: Short talk of 20 mins Session type: Short talk of 20 mins

View proposal in schedule

Abstract

ShareChat is India’s largest vernacular social network platform built to enable next generation of India’s internet users. ShareChat is available in 14 vernacular languages. At ShareChat our data is fresh, with most users coming online for first time, our primary goal is to server most relevant content to the users at appropriate time. In this talk we will discuss the new challenges these first time internet user present. We will motivate the feed generation problem and give a walkthrough of Feed Generation algorithm at ShareChat.

Outline

  1. Introduction to ShareChat
  2. Understanding next billion users
  3. Feed Generation Problem at ShareChat
  4. Building a relevance model
  5. Feed Generation Architecture at ShareChat
  6. Data Challenges: Challenges in designing feed generation for ShareChat and unique insights that ShareChat’s data presents.

Speaker bio

Ayush is currently a lead data scientist at ShareChat. He designs algorithms for content-relevance and feed-generation. Ayush comes with past experience of working on a varied set of data science problems in different domains including Healthcare, Fin-Tech, Life Sciences and Manufacturing domains.

Slides

https://docs.google.com/presentation/d/1q9wcHYiquoKeEI1DkL9P6GZMS4lmwTP8bM3Z259hz2M/edit?usp=sharing

Comments

  • Zainab Bawa (@zainabbawa) Reviewer 4 months ago (edited 4 months ago)

    The permission settings on the slides have to be changed to public, Ayush.

    We’ll need to see the slides to understand the details of what you have mentioned in the abstract.

  • Zainab Bawa (@zainabbawa) Reviewer 4 months ago

    Thanks for making the slides public, Ayush.

    Much of the slides after the intro are about the theory and comparison of deep learning in recommender systems. This makes the talk more a theoretical exposition to deep learning for recommender systems rather than a talk about the feed generation problem at Sharechat. Reduce this portion of the presentation.

    Here are some comments for which we need your responses:

    1. How (and why) do the deep learning algortihms underlying Youtube and Google recommender systems compare with Sharechat’s context and product offering?
    2. It is unclear what is the problem at ShareChat for which you need to use Deep Learning and recommender systems? Explain the problem and the context in detail?
    3. What other approaches and solutions did you consider when solving the feed generation problem before narrowing down on this approach?
    4. What is the one innovation that you consider as a big win for the solution you came up with?
    5. How is the situation different before you implemented this solution vis-a-vis after you implemented this solution?

    Some of the slides are incomplete. These have to be completed by Wednesday, for a full review.

  • Zainab Bawa (@zainabbawa) Reviewer 4 months ago

    Hello Ayush,

    The talk looks much tighter now. Thanks for all the hard work. Couple of points of feedback, mostly on the visual appearance of the slides:

    1. The fonts on the slide are too small, even for a one-on-one scale. The font size has to be increased to the auditorium level of scale.
    2. The visuals also need zooming in, and increasing in size.

    Will share the rehearsal invite with you for pre-event prep.

Login with Twitter or Google to leave a comment