Email: dhruv@umn.edu
Research Motto: In theory, there is no difference between theory and practice. But, in practice, there is.
I am currently a Post-Doc Researcher at Microsoft Research. Prior to this, I completed my PhD under Prof. Abhishek Chandra in the Computer Science Department at the University of Minnesota-Twin Cities. During my PhD, I also collaborated with Prof. Ramesh Sitaraman at University of Massachusetts, Amherst. Broadly, I am interested in Distributed Systems and its application in Data Analytics and Machine Learning. Currently, I am working on problems related to Geo-distributed Analytics and also looking to apply data-driven techniques to optimize systems.
I graduated in 2014 with a Bachelors in Computer Science from BITS-Pilani, India. At BITS-Pilani, I was a part of Advanced Data Analytics and Parallel Technologies Lab, where I worked on parallel and distributed algorithms for data mining.
Selected Publications
HACCS: Heterogeneity-Aware Clustered Client Selection for Accelerated Federated Learning [Paper]
Joel Wolfrath, Nikhil Sreekumar, Dhruv Kumar, Yuanli Wang, Abhishek Chandra
Towards WAN-Aware Join Sampling over Geo-Distributed Data [Paper]
Dhruv Kumar, Joel Wolfrath, Abhishek Chandra, Ramesh Sitaraman
AggNet: Cost-Aware Aggregation Networks for Geo-distributed Streaming Analytics [Paper]
Dhruv Kumar, Sohaib Ahmad, Abhishek Chandra, Ramesh Sitaraman
Accelerated Training via Device Similarity in Federated Learning [Paper] [Video]
Yuanli Wang, Joel Wolfrath, Nikhil Sreekumar, Dhruv Kumar, Abhishek Chandra
Exploiting Data Heterogeneity for Performance and Reliability in Federated Learning [Poster]
Yuanli Wang, Dhruv Kumar, Abhishek Chandra
DeCaf: Iterative Collaborative Processing over the Edge [Paper]
Dhruv Kumar, Aravind Alagiri Ramkumar, Rohit Sindhu, Abhishek Chandra
A TTL-based Approach for Data Aggregation in Geo-distributed Streaming Analytics [Paper]
Dhruv Kumar, Jian Li, Abhishek Chandra and Ramesh Sitaraman
TTL-based Approach for Data Aggregation in Geo-Distributed Streaming Analytics [Poster]
Dhruv Kumar, Jian Li, Abhishek Chandra and Ramesh Sitaraman
Poster at USENIX OSDI 2018
Education
University of Minnesota-Twin Cities, Minneapolis, United States
PhD in Computer Science (September 2017 - Current)
CGPA: 4.0/4.0
3M Science and Technology Fellowship.
Birla Institute of Technology and Science, Pilani, India
B.E.(Hons.) in Computer Science (August 2010 - May 2014)
CGPA: 9.92/10
Rank 1 in Class of 2014 of Computer Science, comprising 120 students.
Rank 3 in Class of 2014 of BITS-Pilani, comprising 800 students.
Work Experience
Google Cloud, Sunnyvale, California
PhD Intern (June 2019 - August 2019)
Project: Learning to prefetch data for disk servers
- Identified an efficient strategy for constructing ground truth and proposed a deep neural network based architecture for predicting the prefetch data.
- Initial evaluation showed an improvement of upto 20% in cache hit rates using the proposed model over the existing approach.
Goldman Sachs, Bengaluru, India
Quantitative Analyst and Software Engineer (November 2014 - April 2016)
- Improved the efficiency of risk-management system by suggesting improvements to the SQL queries going to Sybase IQ database.
- Assisted in migrating from Sybase IQ database to MemSQL database for faster access.
- Implemented a H2-database based server for allowing real-time updates to the tables residing in the servers.
- Learnt about the real life use-cases of databases.
Other Publications
Simulating Landscape Hydrologic Connectivity in a Precise Manner Using Hydro-Conditioning
Rallapalli Srinivas, Matt Drewitz, Joe Magner, Ajit Pratap Singh, Dhruv Kumar, Yashwant Bhaskar Katpatal
In Advances in Computational Modeling and Simulation. Lecture Notes in Mechanical Engineering. Springer, Singapore. 2022
Hydro-climatic optimized machine learning modeling for water quality assessment of a river based on uncertainty and sensitivity analysis
Srinivas Rallapalli, Ananya Jain,Dhruv Kumar
Under submission in Journal of Cleaner Production, Elsevier, 2021.
Grid-R-tree: a data structure for efficient neighborhood and nearest neighbor queries in data mining
Poonam Goyal, Jagat Sesh Challa,Dhruv Kumar, Anuvind Bhat, Navneet Goyal, Sundar Balasubramaniam
In International Journal of Data Science and Analytics (JDSA), Springer, 2020.
An Efficient method for Batch Updates in OPTICS Cluster Ordering
Dhruv Kumar, Poonam Goyal, Navneet Goyal.
Exact, Fast and Scalable Parallel DBSCAN for Commodity Platforms
Poonam Goyal, Sonal Kumari, Ankit Sood, Dhruv Kumar, Sundar Balasubramaniam, Navneet Goyal
In International Conference on Distributed Computing and Networking (ICDCN), 2017
A fast, Scalable SLINK Algorithm for Commodity Cluster Computing Exploiting Spatial Locality
Poonam Goyal, Sonal Kumari, Sumit Sharma, Dhruv Kumar, Vivek Kishore, Sundar Balasubramaniam, Navneet Goyal.
In IEEE International Conference on High Performance Computing and Communications (HPCC), 2016.
Parallelizing OPTICS for Commodity Clusters
Poonam Goyal, Sonal Kumari, Dhruv Kumar, Sundar Balasubramaniam, Navneet Goyal, Saiyedul Islam, Jagat Sesh Challa.
In International Conference on Distributed Computing and Networking (ICDCN), 2015.
Parallelizing OPTICS for multicore systems
Poonam Goyal, Sonal Kumari, Dhruv Kumar, Sundar Balasubramaniam, Navneet Goyal.
In ACM India Computing Conference (ACM COMPUTE), 2014.