Bhuvesh Kumar

Machine Learning Scientist

Biography

I am a Machine Learning Scientist at TikTok where I work on the main recommendation team. I design algorithms to to bring joy and inspiration to hundreds of millions of users every day! Before Tikok, I obtained my Computer Science PhD from Georgia Institute of Technology where I was advised by Prof. Jacob Abernethy and Prof. Jamie Morgenstern. I am broadly interested in machine Learning and algorithmic economics. My PhD work focuses on designing machine learning systems with theoretical foundations operating under the presence of social and economic forces. It spans emerging fields such as mechanism design, robust machine learning, active learning, federated learning, and fairness and privacy in machine learning.

I graduated from IIT Kanpur, India with a bachelors degree in Computer Science and Engineering where I worked with Dr. Prateek Jain and Prof. Purushottam Kar.

Interests

Recommendation Systems
Privacy and Fairness in A.I.
Mechanism Design
Online Learning

Education

PhD in Computer Science, 2022

Georgia Institute of Technology
B.Tech in Computer Science and Engineering, 2017

Indian Institue of Technology Kanpur

Publications

Accelerated Federated Optimization with Quantization

with Yeojoon Youn, and Jacob Abernethy

Oral at Federated Learning Worshop, NeurIPS 2022

Details

ActiveHedge: Hedge meets Active Learning

with Jacob Abernethy, and Venkatesh Saligrama

ICML 2022

Details Proceedings ICML Internal Page poster

Observation Free Attacks on Stochastic Bandits

with Yinglun Xu, and Jacob Abernethy

NeurIPS 2021

Details NeurIPS

Optimal Spend Rate Estimation and Pacing for Ad Campaigns with Budgets

with Jamie Morgenstern, and Okke Schrijvers

Talk at INFORMS Annual Meeeting 2021

Details arXiv talk

Bridging Truthfulness and Corruption-robustness in Multi-Armed Bandit Mechanisms

with Jacob Abernethy, Thodoris Lykouris, and Yinglun Xu

Incentives in Machine Learning, ICML 2020

Details talk

Learning Auctions with Robust Incentive Guarantees

with Jacob Abernethy, Rachel Cummings, Jamie Morgenstern, and Samuel Taggart

NeurIPS 2019

Details pdf NeurIPS Poster

Accelerated Parallelizable Projection-Free Algorithm for the Nuclear-Norm Ball Constraint

with Jun-kun Wang, and Jacob Abernethy

in submission

Details

Teaching

Graduate Teaching Assistant, Georgia Tech

CS7646, Machine Learning for Trading (Spring 2022, Summer 2022)
CS7545, Machine Learning Theory (Fall 2018, Fall 2019)

Tutor, IIT Kanpur

ESC101, Fundamentals of Computing (Fall 2016, Spring 2017). Delivered lectures for the online version of the course as well.

Activities

Georgia Tech

I was the founding vice-president of the newly formed School of Computer Science Graduate Student Association for the 2021-2022 academic year.
I was the student nominee for faculty hiring at the School of Computer Science at Georgia Tech for the 2019-2020 academic year. I organized student meetings for the faculty candidates and represented the student body in faculty hiring decisions.
I used to co-organise the Georgia Tech ACO student seminar.
We have a reading group on machine learning theory.
I co-organise the College of Computing Grad Student Happy Hour. Join us on our facebook group or mailing list.

IIT Kanpur

Programming Club, IITK: I was the coordinator of the programming club of IIT Kanpur in 2015-16.
Science Coffeehouse, IITK: I was one of the organizers of the Science Coffeehouse in 2015-16, the science and mathematics discussion/talks group at IIT Kanpur.