Feb 2023 – Apr 2024 · 1 yr 2 mos · Dallas, Texas, United States · Remote

• Lead outbound fulfillment services team at GoPuff, tasked with efficiently delivering essential goods to customers. This spans entire packing lifecycle, from order placement to handover for dispatch by driver.
• Led multiple XFN projects ranging from reducing RTT latency, minimizing packing errors, managing order substitutions, order prioritizations, optimizations and enhancing various service features.
• Managed ~7 different micro-services, driving efficiency improvements, features, focusing on availability and latency and enforcing strict SLAs with real-time alert capabilities.
• Conceptualized, designed and worked hands on with team for pack time estimation service improving RTT of customer delivery time by ~30%.
• Designed and worked XFN on reducing mispacks from ~5% to ~0.25%.

May 2019 – Feb 2023 · 3 yrs 9 mos · Menlo Park, California, United States · Hybrid

• Worked on distributed batch scheduler running ~100M jobs per day constituting of machine learning, data pipeline, oculus and all generic Meta workload.
• Led team of 8 dedicated to orchestrating job management stack platform.
• Worked extensively on various distributed problems including improving releases, SLAs/SLOs, sharding, scaling, load testing, database and performance issues.
• Designed and led multiple long term XFN collaboration projects enhancing features and integrating with various teams in and outside org.
• Organized a long term core schedulers summit through out Meta to onboard various workload onto our system.
• Number of actively running jobs grew from ~10M to ~100M during tenure.

Sep 2015 – May 2019 · 3 yrs 8 mos · Newark, California, United States · On-site

• Worked on broad range of performance enhancement features covering almost all layers in ZFS file system based on Solaris, especially on improving raid space bloat issue and recompaction engine for same.
• Designed and implemented VMware vvol targets on top of ZFS.
• Granted with 2 patents and internal paper for same.

Jul 2014 – Sep 2015 · 1 yr 2 mos · On-site

• ESX Space Loaner
• This turns the ESX server into a storage target with a target FC driver, SCST module to expose its underlying local SSDs and/or remote SSDs via fabric to loan some SSD storage space to other Hypervisor servers. This serves as a huge benefit for the vSAN.
• Data Security
• Designed and implemented a solution for providing vmdk level encryption with centralized key management server.
• Research as a Service
• Worked in design and implementation of RAAS which is a self-service portal for researchers to collaborate, compose and run research algorithms on distributed, secured and compliant datasets.

Feb 2013 – Jul 2014 · 1 yr 5 mos · On-site

• Dif-Dix
• Worked on providing DIF (data integrity field) and DIX (Data Integrity Extension) support to vsphere. DIF provides the ability to protect data from the HBA till the time the I/O is written to disk and DIX allows this protection to extend all the way from the OS to the HBA.
• IO-Hinting
• Worked in collaboration with Brocade to provide hints and coordinate Service Level Agreements for the underlying infrastructure for making the application aware of the underlying storage infrastructure. Published a paper in VMware’s internal technical journal
• (Link: https://labs.vmware.com/vmtj/application-aware-storage-platform).

Jun 2012 – Aug 2012 · 2 mos · Palo Alto, United States

• Worked in EE-Storage Development R&D team on the design and development of Cache Coherency model. Currently both Host and Target are caching the I/O blocks which causes redundant copies of blocks to exist in both caches. I worked and implemented a solution in order to avoid this problem.
• My Intern Work involved:
• 1. Studying and understanding the ESX Storage Stack.
• 2.Working on the I/O infrastructure to pass hints to storage array.
• 3. Integrating the new I/O infrastructure with vFlashCache (SSD Caching) Layer of ESX server.
• SCSI protocol is used as the communication channel. Due to this model a lot of cache space is preserved and many significant benefits can be achieved in terms of space and performance. It is expected to standardize this model with International T10 Committee.

Dec 2011 – Apr 2012 · 4 mos

• Working on project "Video messages with QR Code for gift items" for E-commerce website with Magento tool.
• Worked on Magento for building a plugin to support additional features of uploading videos through E-Commerce websites and storing them on the cloud.
• Now Working with Servlets and testing team.

Aug 2011 – Dec 2012 · 1 yr 4 mos

• Doing my Masters in Computer Sciences.