Unlocking Personalization in Large-Scale Recommendations

Abstract
In this talk, we delve into the fascinating world of large-scale recommender systems, exploring three key dimensions: Conformity, Exploration and Context Awareness. We present innovative approaches to break free from conformity bias, infuse exploration to capture evolving user interests and incorporate context awareness for more personalized recommendations. Through rigorous experimentation and real-world deployments on platforms like Facebook Watch, we unveil how these methodologies revolutionize recommendation systems. Join us to uncover the secrets behind unlocking personalization in large-scale recommendations, offering profound insights that bridge the gap between user engagement and system performance.
Speakers

• Khushhall Chandra Mahajan is a Senior Machine Learning Engineer and Researcher working in recommendation systems. Presently, he is an integral part of the Video Recommendations team at Meta, where he plays a pivotal role in crafting cutting-edge ML algorithms. These algorithms directly enhance the accuracy and impact of video recommendations used by billions of users worldwide. His work powers the recommendation engine for Facebook Watch and Reels, reaching over 1.25 billion users monthly. Prior to this role, he was in the Ads team, developing innovative Ads solutions that drove top-line revenue for the company. He has published several research papers in the domain of machine learning and recommendation systems. His research interest focuses on exploration & diversity in recommendation system and realtime ranking. Additionally he has served on the program committee of various top-tier international machine learning conferences such as NeurIPS, ICLR, ECIR, AISTATS, etc and organizer of the VideoRecsys workshop in ACM RecSys conference. As a fascinating side note, his innovation extends beyond machine learning; he has also developed the Swarachakra Bangla Android keyboard, which has garnered over a million downloads.

• Amey Porobo Dharwadker works as a Machine Learning Engineering Manager at Meta, where he leads the Facebook Video Recommendations Core Ranking team responsible for developing personalization models used by billions of users around the world. His work has been instrumental in driving the remarkable growth in active users for Facebook, by powering the success of Facebook Watch and Reels, reaching more than 1.25 billion monthly users. Prior to that, he delivered significant user engagement and revenue growth for Facebook through advancements in News Feed and Ads Machine Learning. He has published several research papers in the fields of large-scale recommendation systems and machine learning. He also actively participates as a program committee member for top-tier AI venues including AISTATS, AAAI, IJCAI, ECIR, CIKM, etc. and organizes the VideoRecsys workshop at ACM RecSys conference. He also serves on the juries of renowned global technology competitions, including the Edison Awards and Globee Information Technology Awards.

Directions to the venue:
SOE/SCIS2 SR B1.2 is located at the basement concourse (B1), which is accessible to the public and requires NO registration with the security. Once you are in the basement concourse, follow signatures towards School of Economics / School of Computing and Information Systems 2.
By train: Alight at either Bencoolen (Downtown Line) or Bras Basah (Circle Line) stations.
- Bencoolen (Downtown Line), please kindly follow the signage towards Exit C, and then take a right turn at the basement concourse (B1).
- Bras Basah (Circle Line), please kindly follow the signage towards Exit B, and then take a right turn at the basement concourse (B1).
By taxi: The drop off point is at 90 Stamford Road, Singapore Management University, Singapore 178903. Then, take the escalator towards the basement concourse.
By driving: Parking is available at the Lee Kong Chian School of Business. Exit the car park via lift lobby A, which is connected to the basement concourse (B1).