United States

Back-end

Data engineering

Architectural Engineering

Security Engineering

Site Reliability

Udbhav Prasad

Staff Software Engineer

Problem-solving

Critical-thinking

Cloud computing

Data mining

Database management

Information architecture

Mathematics

Python

Docker

GitHub

Golang (Go)

Greenhouse

Java

Jupyter Notebook

Lucidchart

NLTK

Notion

Numpy

gRPC

REST API

PostgreSQL

Linear

– Near-zero Recovery Point Objective on backup performance over a Network File System on dozens of Oracle platforms and configurations implemented across Rubrik enterprise customers.
– Designed and implemented the Live Mount and Export data recovery workflows to recover production databases, offering near-zero Recovery Time Objective and highly customizable point-in-time copies of Oracle database backups.
– Designed and implemented backup and recovery of NetApp NAS volumes with their ONTAP REST API. Laid the foundations for Rubrik’s partnership with NetApp, enabling first-to-market support for NetApp’s proprietary SnapDiff API.

• Contributed extensively to building a replication micro-service that copies data efficiently from the minimum number of nodes in a source Cassandra cluster to a target Cassandra cluster.
• Member of data platform team responsible for developing and maintaining globally available Cassandra cluster running on multiple AWS regions along with services and applications that support scaling data access for both frontend applications and backend pipelines
• Launched globally available Cassandra cluster across AWS on East, West and APAC datacenters.
• Horizontally scaled Cassandra backend infrastructure at low cost by leveraging AWS spot instances.

• Working on and managing an Elastic Data Platform that provides infrastructure as a service - provisioning AWS instances, scaling up and scaling down instantly and handling optimal resource allocation.

• Worked on various projects that involved cutting infrastructure costs without affecting our serving abilites at scale. This included networking throttling techniques at application and system levels to use system resources more effectively during data transfers, working on the Elastic Data Platform to improve resource allocation for data pipelines.

• Worked on improving the scalability of the Cassandra-based distributed database system without horizontal or vertical scaling.
• Cassandra, although good for data-modeling elasticity, is not famous for it's clustering elasticity. That is, it is difficult to add or remove nodes at will, especially when the existing clusters have a lot of data in them. We also have a different number of pipelines writing to and reading from our Cassandra clusters, with very high QPS at peak times. It is infeasible to maintain a Cassandra cluster to address the peak-time traffic requirements.
• In order to not overwhelm the cluster, and to maintain a small cluster size, we have BC2, which provisions spot-clusters as pipelines demand it, and replicate only the data required from the main cluster onto this spot cluster. The pipelines can now run amok on the "BC2 cluster" as we call it, reading and writing on the snapshot of the data. This takes care of the read pressure on the main cluster. However, in order to keep the data up-to-date, all the writes need to be propagated from the BC2 cluster to the main cluster.
• As it was, we were writing the data back using Cassandra's thrift client. During my internship, I explored and compared two techniques: simply copying the SSTables from the BC2 cluster to the main cluster and doing a refresh, versus using Cassandra's native BulkLoader.

• Used web server log data to detect and monitor competitor activity, which was then fed to the Walmart Pricing team.
• Worked with Hadoop and Hive on the vast amount of log data at our disposal.
• Implemented a complete pipeline beginning with the log input, and ending with a summary of potential competitors and reporting of other suspicious activity.
• Wrote a framework in Python to implement this pipeline. The framework was completely modular, and each segment of the pipeline was customizable in a plug-in sort of way.
• Wrote Google Chrome browser extensions in JavaScript to improve the user experience on the team’s dashboard, which is used by merchants who interact with the team and imbibe our data.
• Implemented the extensions atop an MVC framework with a backend in Ruby on Rails.

Udbhav Prasad🇺🇸

Create an account to accelerate your career with expert mentors worldwide.