Oct 2022 – Apr 2025 · 2 yrs 6 mos

Dec 2021 – Oct 2022 · 10 mos

Mar 2019 – Dec 2021 · 2 yrs 9 mos

Jan 2017 – Mar 2019 · 2 yrs 2 mos

Jul 2014 – Dec 2016 · 2 yrs 5 mos · Bengaluru, Karnataka, India

• Email Time Prediction:
• Ebay marketing strategy involves notifying the users through more than 30 different
• emails about item price reduction, watch item ending, Out Bid Notification etc. The
• amount of revenue generated is directly proportional to open and click rates. So the
• email sending time really matters and it has been justified through several researches as
• well. Most recipients (95 percent) open and read the email within the first 60 minutes
• and if the email goes unopened for 24 hours, the chances of it being read come close
• to zero. The idea is to send email to user during his peak opening time to increase
• open and click rates. Developed an algorithm to model user email opening behavior
• and predict most favorable time for sending emails.
• See It Want It Get It:
• One big fallback every e-commerce user experiences is not able to find what they are
• looking for. This is mainly because they cannot express their thoughts with words.
• Developing an Image searching app is becoming necessary for any e-commerce company.
• This idea is worked on after office hours and developed a mobile app from which a picture
• can be taken and user will be directed to item page on eBay. For this daunting project
• developed a Custom feature extraction method based on Deep Belief networks and a
• custom dimensionality reduction algorithm , which can index images for similar item
• retreival. The accuracy of algorithm over all of eBay.in(around 30Tb of data) is 100
• percent in case of perfect match and highly satisfactory in case of similar searches. This
• project won at Skunkworks which is Global innovation challenge for eBay.

May 2013 – Jul 2013 · 2 mos · Bengaluru, Karnataka, India

• Publisher Fraud Detection:
• As a part of their advertising campaign Amazon bids for the online spaces of various
• websites. The bidding amount for a website is influenced mainly by click through rates
• and conversion rates of that particular website. Recently botnets like chameleon have
• been developed which infect users systems and divert traffic to some websites which will
• result in loss of huge share of advertising costs. We successfully developed a model to
• differentiate such faulty websites from genuine websites leading to frugal utilization of
• advertising resources. Ran this model for around 200 million users on Hadoop cluster.

May 2011 – Jul 2011 · 2 mos · Bengaluru, Karnataka, India

• Internship at Qualcomm Inc. in Windows Phone Development team for a period of two months.
• Asynchronous parallel testing of Multiple Mobiles:
• Developed hardware and software for testing multiple windows based mobile phones
• simultaneously. Impact of the project is cutting off the time and effort of quality testing
• by more than 75 percent.