Nov 2023 – Present · 2 yrs 4 mos · San Francisco Bay Area · On-site

• myaidrive.com

Aug 2018 – Nov 2021 · 3 yrs 3 mos · San Francisco Bay Area

• Google Cloud AI team.
• Current focus: TensorFlow for Enterprise, and https://tensorboard.dev
• Past: https://github.com/kubeflow/fairing

Jan 2018 – Aug 2018 · 7 mos · San Francisco Bay Area

• Doing my part in making autonomous vehicles a reality!
• Working on distributed training and scalable inference.

Feb 2017 – Jan 2018 · 11 mos · San Francisco Bay Area

• Leading a team of awesome data scientists focused on solving fraud and security problems using machine learning. We apply cutting edge deep learning techniques like sequence modeling to model users' behavior. Dealing with one of the unique and large datasets out there.

Jan 2016 – Feb 2017 · 1 yr 1 mo · San Francisco Bay Area

• Keeping your Uber account safe from hackers!

Sep 2014 – Jan 2016 · 1 yr 4 mos · San Francisco Bay Area

• Built a SaaS product for anomaly detection on KPIs and other business metrics.
• Built an IoT device for monitoring driver behavior and crash detection in taxis.
• Built a real-time in-memory columnar analytic engine for log analysis that is capable of querying several GBs of in-memory data per second per core.
• Designed and built a columnar in-memory format for JOSN that is several times faster and compressible than the standard row format.

Mar 2013 – Sep 2014 · 1 yr 6 mos

• Responsible for keeping the LinkedIn members safe from account takeovers. Built many data products to defend against humans and bots that are trying to take over members' accounts. While doing that, also made many improvements to reduce friction for members logging into their own account.
• Architecting an internal tool that needs to store and index large volumes of tracking data.
• PS: It's hard to give details when you work on security products :)

Aug 2011 – Feb 2013 · 1 yr 6 mos

• Built IP/ISP/Email reputation based on LinkedIn and external data.
• Developed new ways to protect LinkedIn accounts from being hacked.
• Responsible for analyzing performance and usage of security products by LinkedIn members.
• Additionally taking care of security incident response and training new members of the security team on LinkedIn data infrastructure and incident response.

Jul 2010 – Jul 2011 · 1 yr

• Handled many security incidents, some involving huge botnets. Built login challenges, like captcha, to protect account take overs. Performed product analytics for many security products. Developed LinkedIn's first real time monitoring system based on Kafka that handle up to half a million events per node.

May 2010 – Sep 2014 · 4 yrs 4 mos

• Hacking the most creative and controversial hacks at LinkedIn!
• Hackday Projects:
• People You Should Know,
• BugIn (Best Production Tool),
• Jobs For Vets,
• InCorrect (Best Friction Reduction),
• inFair,
• IN or OUT (Most Popular),
• Klunk (Best Internal Tool),
• Are you INterested? (Most Controversial),
• FollowIn,
• Holy Grail (Most Controversial Project)

May 2010 – Jul 2010 · 2 mos

• In short time, learnt LinkedIn's big data infrastructure and built first profile scoring model at LinkedIn to distinguish between fake and real members. Additionally built a model to detect profanity in profiles.

Jan 2009 – Dec 2009 · 11 mos

• Fall 2009: Analysis of Algorithms-I, Instructor: Prof. Clifford Stein
• Spring 2009: Data Structures in Java, Instructor: Prof. Alexander Pasik

Jan 2009 – May 2009 · 4 mos

• Intrusion Detection Systems Lab:
• Designed a distributed system for sharing of alerts and traffic models among collaborating sites while preserving.
• Developed correlation techniques to correlate IP alerts while preserving privacy using bloom filters
• Involved in developing a new feedback technique for use in anomaly detection sensors that uses partial labels of the training data
• Designed a web-crawler for mining error pages in a website

Jun 2007 – Jun 2008 · 1 yr

• Involved in RTL synthesis for Sharc processor and optimizing architecture for the target speed. Additionally, developed tests for media controller for the Sharc processor.

Jan 2007 – Apr 2007 · 3 mos

• Built a source code search engine for HP's internal repository that is aimed to facilitate code reusability. Developed algorithms to parse structure in source codes of different programming languages and built an intuitive web application for end users.