Sep 2016 – Jun 2024 · 7 yrs 9 mos · San Mateo, California · On-site

• Rockset is a serverless realtime search and indexing database that delivers millisecond-latency SQL on raw, semi-structured data, without any ETL. Rockset enables developers to build decision-making applications like personalization engines, gaming leaderboards, IoT apps and vision AI-based automation on massive datasets.
• Usage in OpenAI: https://www.linkedin.com/feed/update/urn:li:activity:7209964394204524545/

Jan 2014 – Jan 2016 · 2 yrs

Jan 2013 – Jan 2016 · 3 yrs · Santa Clara, California

• Acquired by HP in 2018

Jan 2012 – Jan 2016 · 4 yrs · Solana Beach, California

• Acquired by Teradata in 2017

Jan 2012 – Jan 2016 · 4 yrs · Mountain View, California

Jan 2012 – Jan 2016 · 4 yrs · Palo Alto, Califoria

Jan 2008 – Jan 2016 · 8 yrs · Menlo Park, California

• I was the founding engineer of the RocksDB embedded open-source database (http://rocksdb.org/). My responsibilities included design and development of core database features; providing performance numbers on varied hardware; capacity planning of database tiers, providing customer support, etc. My introductory talk about RocksDB is at https://www.youtube.com/watch?v=V_C-T5S-w8g.
• Prior to this, I contributed core features for Apache HBase and was instrumental in the deployment of HBase for various applications at Facebook.
• Earlier, I was instrumental in scaling Facebook's Apache Hadoop File System's clusters to tens of petabytes. I designed and implemented core features of HDFS including distributed encoding, federation and high-availability that were deployed internally at Facebook.
• Some of the services that I have helped design, develop and deploy at Facebook are a low-latency data publishing service for warehouse data, a storage service especially designed for storing small messages, a large-scale and high fanout publish-subscribe system for distributing small pieces of data across multiple geo-locations, a distributed-encoding-enabled storage system for storing photos, etc.

Jan 2006 – Jan 2017 · 11 yrs

• https://issues.apache.org/jira/projects/HDFS/summary

Jan 2006 – Jan 2008 · 2 yrs · Santa Clara, California

• I was the founding architect and engineering lead of the Apache Hadoop File System (HDFS) and a member of the Apache Hadoop Project Management Committee (PMC). I designed and implemented some of the core features of HDFS, provided customer support to internal customers, published benchmark results and helped recruit people into the HDFS team at Yahoo. I wrote the initial version of the HDFS Architecture here https://hadoop.apache.org/docs/r1.2.1/hdfs_design.pdf.

Jan 2004 – Jan 2006 · 2 yrs · Milpitas, California

• Mendocino Software was a startup focussing on continuous data protection systems. I was a team lead of a 6 people team and was responsible for development of the core engine. I played a critical role in a collaborative engineering development process with engineers from EMC Corporation.

Jan 1998 – Jan 2004 · 6 yrs · Mountain View, California

• I developed core components of the Veritas File System (VxFS). This needed expertise in kernel-mode drivers, vfs layers and OS-page cache behaviour on SunOS, IRIX, AIX and HP. I developed the first port of VxFS to WindowsNT.

Jan 1998 – Jan 1998 · 0 mo · San Jose, California

• Oreceipt allows you to keep track of your invoices, organize them in folders of your choice, print your invoices anytime you desire, forward individual or groups of invoices by e-mail, or access your data directly in an XML format. This was a four person company; I was a co-founder and was reporting to the CEO.
• I was responsible for software design and development for managing electronic receipts issued by retailers. These needed software integration with various point-of-sale software; their transmission to servers managed by oreceipt.com; and the access, retention and display of these receipts by businesses and consumers. This software was acquired by the Adelman Travel Group at a later date and still serves real customers (https://www.adelmantravel.com/Reporting-Tools-About-Oreceipt.asp).

Jan 1995 – Jan 1998 · 3 yrs · Greater Pittsburgh Region

• I wrote code for the Andrew Distributed File System (AFS). This was the world's first distributed file system and was widely used by large corporations with a federation of AFS cells distributed across geo-locations. I was responsible for developing core components of the fileserver, the vlserver and the ubik quorum election protocol. I provided onsite customer support for high-profile customers like CERN, Intel, etc.

Aug 1991 – Jul 1993 · 1 yr 11 mos · Greater Delhi Area

• CDOT is a premier research and development center in India with a focus on development of digital switching systems. I developed code for traffic, routing and billing of calls in a distributed system that included the RAX and the MAX. These systems revolutionized the nascent Indian telecom industry within a span of few years.