Oct 2022 – Dec 2022 · 2 mos

Aug 2022 – Dec 2022 · 4 mos

Dec 2021 – Aug 2022 · 8 mos

Feb 2020 – Jan 2022 · 1 yr 11 mos

Sep 2019 – Feb 2020 · 5 mos

• Learning Optimizer
• DBA-as-a-Service

Apr 2018 – Aug 2019 · 1 yr 4 mos

• Learned cardinality models in SCOPE query optimizer.
• CloudViews for Spark SQL workloads in HDInsight.
• Workload optimization platform for cloud query engines.
• Publications:
• Towards a Learning Optimizer for Shared Clouds, VLDB 2019.
• SparkCruise: Handsfree Computation Reuse in Spark, VLDB 2019.

Aug 2016 – Mar 2018 · 1 yr 7 mos

• Built CloudViews for automatic computation reuse in a massive-scale cloud service at Microsoft. Developed the system and a scalable algorithm for subexpression selection.
• Publications:
• Selecting Subexpressions to Materialize at Datacenter Scale, VLDB 2018.
• Computation Reuse in Analytics Job Service at Microsoft, SIGMOD 2018.

Aug 2015 – Jul 2016 · 11 mos

• Analyzed production workloads from Microsoft's SCOPE query optimizer.
• Combining resource and query optimizations in Spark and Hive.
• Publications:
• Query and Resource Optimization: Bridging the Gap, ICDE 2018
• Query and Resource Optimizations: A Case for Breaking the Wall in Big Data Systems, arXiv:1906.06590

Apr 2013 – Aug 2015 · 2 yrs 4 mos · Cambridge, USA

• Graph Analytics using Relational Databases: Developed techniques for efficient in-database graph analytics. Specifically, I looked at four things (i) comparing different systems for graph analysis, (ii) query optimizations techniques to tune the performance of graph queries on the column-oriented, (iii) ease-of-use of doing graph analytics in relational databases; proposed two alternatives, a vertex-centric query interface in SQL and a new Pig-like language called GRAPHiQL, (iv) providing better support for iterations by loading graphs into an optimized data structure in main memory.
• Data Preparation & Cleaning: Built Cartilage to allow developers to have full control over data preparation in a distributed file system, such as HDFS. Cartilage provides a dataset abstraction between the logical user dataset and the actual physical HDFS files. We introduced data plans, analogous to query plans, to enable users to declaratively specify their transformations, from the logical dataset to the physical HDFS files.
• Publications:
• GraphiQL: A Graph Intuitive Query Language for Relational Databases, IEEE BigData 2014.
• Graph Analytics using Vertica Relational Database, IEEE BigData 2015.
• A Robust Partitioning Scheme for Ad-Hoc Query Workloads, SOCC 2017.
• AdaptDB: Adaptive Partitioning for Distributed Joins, VLDB 2017.

Dec 2012 – Mar 2013 · 3 mos · Saarland

• Database Cracking: identified several limitations of existing cracking algorithms, proposed novel algorithms to improve the convergence and robustness of database cracking, introduced fingerprinting of different indexing methods, and offered promising directions for future research.
• Publications:
• The Uncracked Pieces in Database Cracking, VLDB 2014.
• An Experimental Evaluation and Analysis of Database Cracking, The VLDB Journal, August 2015.

Mar 2007 – Sep 2008 · 1 yr 6 mos · Gurugram, Haryana, India

• Developed vertical search engines using Lucene/Solr technologies and their front-end using LAMP. Built web analytics framework to analyze search engine activity, including user behavior, search
• engine performance, and search quality, on a daily as well as real-time basis. Prototyped a lead-generation system for Mobile Ads Business.

Jul 2006 – Feb 2007 · 7 mos · Bengaluru, Karnataka, India

• Worked on building an order fulfillment system for the Mobility division.

May 2005 – Jul 2005 · 2 mos · Pune/Pimpri-Chinchwad Area

• Implemented an autonomic logging and recovery infrastructure using network blocked devices.