I am a Ph.D. student advised by Professor Raul Castro Fernandez in the ChiData group. My recent work aimed to execute SQL workloads across cloud databases to reduce overall monetary cost. I am also interested in how query optimization and data management intersects with notions of the value of data.
Saving Money for Analytical Workloads in the Cloud
Executing SQL workloads in the cloud across multiple databases, each using a different cloud pricing model, so that overall workload cost is reduced.
For example, BigQuery charges $6.25/TB scanned while Redshift charges about $1/hour for a 1-node cluster. A query scanning 30GB and running for an hour is much cheaper in BigQuery than Redshift.
Published in VLDB 2024
Penelope: Peer-to-peer Power Management
A decentralized approach to power management in large clusters. Clusters have system-wide power caps they must remain under to avoid damage, however intelligent allocation of power to individual nodes can have a dramatic influence on performance.
Nodes communicate with each other to distribute power in a cluster rather than relying on a single coordinator node, creating a more robust and scalable protocol.
Published in ICPP 2022
Conference
Saving Money for Analytical Workloads in the Cloud VLDB 2024
Tapan Srivastava, Raul Castro Fernandez
Penelope: Peer-to-peer Power Management ICPP 2022
Tapan Srivastava, Huazhe Zhang, Henry Hoffmann
Posters
Saving Money for Analytical Workloads in the Cloud, GCASR: Greater Chicago Area Systems Research Workshop
Illinois Institute of Technology, April 2023
Saving Money for Analytical Workloads in the Cloud, CERES Summit
University of Chicago, May 2023
Bauplan
Jun. 2024 - Sep. 2024
Ph.D. Intern
Uber ATG
Jun. 2019 - Sep. 2019
Software Engineering Intern
Built debugging and logging tools for tablets onboard the self-driving vehicle via gRPC servers and protobuf.
Uber ATG
Jun. 2018 - Sep. 2018
Software Engineering Intern
Designed, implemented, and deployed a new backend service built with Go, Thrift, Kafka, and gRPC into production.
University of Chicago, PhD in Computer Science
Sep. 2020 - Ongoing
Design of a method to build a probabilistic generative model
Development of model calibration algorithm for cross-domain few-shot learning
University of Chicago, MS in Computer Science
Sep. 2020 - May 2023
Advised by Prof. Raul Castro Fernandez
Master’s Paper: Saving Money For Analytical Workloads in the Cloud
University of Chicago, BS Honors in Computer Science
Sep. 2016 - Jun. 2020
GPA: 3.9/4.0
Bachelor’s thesis advised by Prof. Hank Hoffmann
University of Chicago, BS in Mathematics
Sep. 2016 - Jun. 2020
GPA: 3.88/4.0
