$$$
{{ $t($store.state.user.experience_value_in_dollars) }}
Senior
{{ $t($store.state.user.experience_search_name) }}
0
jobs
Senior Software Development Engineer at Amazon
Piyush Yawalkar
,
London, United Kingdom
Experience
Other titles
Skills
I'm offering
I have 7+ years of industry experience working in companies like Google, Facebook and Amazon
Markets
United Kingdom
Links for more
Once you have created a company account and a job, you can access the profiles links.
Language
English
Fluently
Ready for
Larger project
Ongoing relation / part-time
Full time contractor
Available
My experience
2014 - ?
job
Software Engineer
Ads Brands.
Projects and Achievements (In Reverse Chronological Order)
○ Video Re-targeting
* Developed the product using which Advertisers target their ads to the users who
watched their videos on Facebook.
* In 4 weeks of soft launch to 0.5% of big brand advertisers, it showed 20% decrease
in cost per conversion over normal post ads.
○ Ads Engagement Re-targeting
* Developed the product for advertisers to re-target their new ads, to users who had
seen or clicked on their previous ads..
○ Video Ads Polling
* Facebook uses polling mechanism to measure ad recall, purchase intent etc and measure the success of an ad-campaign. In this project, the polls are now shown to
the users after they watch a video.
* This results in 7-10% higher recall because watching a video shows higher interest
and remembrance.
○ Video Re-targeting
* Developed the product using which Advertisers target their ads to the users who
watched their videos on Facebook.
* In 4 weeks of soft launch to 0.5% of big brand advertisers, it showed 20% decrease
in cost per conversion over normal post ads.
○ Ads Engagement Re-targeting
* Developed the product for advertisers to re-target their new ads, to users who had
seen or clicked on their previous ads..
○ Video Ads Polling
* Facebook uses polling mechanism to measure ad recall, purchase intent etc and measure the success of an ad-campaign. In this project, the polls are now shown to
the users after they watch a video.
* This results in 7-10% higher recall because watching a video shows higher interest
and remembrance.
Video, Facebook, It, Software, Campaign, ADS
? - ?
job
Manager
Adrian Ford.
I invented a strategy to do the testing and integration of Prime Video Application on Living
Room Devices using AWS IoT, which simplified the solution to not rely on external test
houses; or have other servers to connect and control the devices to drive testing on debug
devices. This solution was reviewed and appreciated by Principal Engineer and Senior
Leaders, and currently in the process of taking the shape of PRFAQ
○ I have a patent filed and pending for this work P62288-US01
● Test Decomposition (Manager: Adrian Ford)
○ I led the Testing Strategy Bar Raiser group which involved 5 orgs, in which I drove the idea
of Test Decomposition; where I suggested to decompose the End-to-end tests in favor of unit/integration tests upstream and run them during the porting process
○ I drove this idea with the help of Proof of Concept. The org is aligned with this idea, and has
led to major initiatives and projects in various teams
● AVTS (Manager: Simon Ortizsi (ortizsi@))
○ Lead the technical design of the Automated Process to create and manage Test Cycles for Living Room Device Manufacturers, through Partner Central and worked with the team to
deliver it.
● Partner Central (Manager: Karan Kapoor (kapoork@))
○ Lead technical design and founder of the tool called Partner Central. Inspired the org for the need of this tool by making mocks and driving through the device approval workflow
○ Inspired the team to move to React for UI development and TestRail for test results
management and had a big role in driving towards it, using spikes and POCs
○ Did the technical design of the Login System, Permissions System, System architecture of the Partner Central Bus aka Platform, Disaster Recovery Plans, Legal Documents
Room Devices using AWS IoT, which simplified the solution to not rely on external test
houses; or have other servers to connect and control the devices to drive testing on debug
devices. This solution was reviewed and appreciated by Principal Engineer and Senior
Leaders, and currently in the process of taking the shape of PRFAQ
○ I have a patent filed and pending for this work P62288-US01
● Test Decomposition (Manager: Adrian Ford)
○ I led the Testing Strategy Bar Raiser group which involved 5 orgs, in which I drove the idea
of Test Decomposition; where I suggested to decompose the End-to-end tests in favor of unit/integration tests upstream and run them during the porting process
○ I drove this idea with the help of Proof of Concept. The org is aligned with this idea, and has
led to major initiatives and projects in various teams
● AVTS (Manager: Simon Ortizsi (ortizsi@))
○ Lead the technical design of the Automated Process to create and manage Test Cycles for Living Room Device Manufacturers, through Partner Central and worked with the team to
deliver it.
● Partner Central (Manager: Karan Kapoor (kapoork@))
○ Lead technical design and founder of the tool called Partner Central. Inspired the org for the need of this tool by making mocks and driving through the device approval workflow
○ Inspired the team to move to React for UI development and TestRail for test results
management and had a big role in driving towards it, using spikes and POCs
○ Did the technical design of the Login System, Permissions System, System architecture of the Partner Central Bus aka Platform, Disaster Recovery Plans, Legal Documents
It, Founder, Manager, LED, Patent, Workflow, Testing, Development, Architecture, Design, Iot, Management, Test, Integration, UI, AWS, Video, React
? - ?
job
Lead the ANVIL review and VIP/OPF onboarding
Onboarding and Operational Runbooks.
along with launching the tool externally
○ Also lead the delivery of other features like Audit Trail
○ I have a patent for the Authentication and Authorization System for this project P43965-US
● ToyBox (Manager: Karan Kapoor (kapoork@))
○ Launched the internal tool ToyBox after joning the team
○ Lead the technical design and implementations of features like export functionality, pagination, features and fields for devices; Regions extension and ANVIL review
○ Also lead the delivery of other features like Audit Trail
○ I have a patent for the Authentication and Authorization System for this project P43965-US
● ToyBox (Manager: Karan Kapoor (kapoork@))
○ Launched the internal tool ToyBox after joning the team
○ Lead the technical design and implementations of features like export functionality, pagination, features and fields for devices; Regions extension and ANVIL review
Design, Audit, Onboarding, Audit, Patent, Manager
2012 - 2014
job
Anti Abuse Engineer in Ads Review Team
Google India.
Projects and Achievements (in reverse chronological order)
○ Similar text Detector
* Google Ads Review System is able to detect normal words, but spamers sometimes
use special characters to trick the review system. Eg AK47, @K47, where @K47
may be missed by review system and their ML models.
* This project enabled to find the closer valid words, using the similar text detection
infrastructure used at Google.
* Now the bulk review tools can capture these bad words and review these ads
○ Mobile Malware Solutions
* Platform which takes URL as input, detect and download the apk files in it and extract features from them.
* These features are then fed to ML models which detects malware in these binaries.
○ Image Ads Clustering
* Continuous pipeline which takes bunch of images as input, extract features from them, and then group the images with similar features together
* Used existing image feature extractors at Google.
* This enabled clustering and bulk reviews of the image ads.
○ Un-reviewed Ads Periodic Pipeline
* Created a continuous pipeline that runs over the un-reviewed ads, and generates
clusters from them using different types of clustering mechanisms.
* New clustering methods can be easily hooked into this infrastructure
* These clusters can now be reviewed using bulk review tools
○ Offline Machine Learning Infrastructure (OLIO)
* A generic infrastructure to create Machine Learning Models for any types of features and labeled data.
* Got First Prize for OLIO in Poster Presentation in Google Anti Abuse Summit 2014, where more than 50 teams participated.
* Used existing command line machine learning tools and created a generic
infrastructure using it
○ AdsDedup
* Project to identify duplicates in un-reviewed ads and predict the review decisions for them, based on the same ads seen previously
* These predictions act as proxy to review the clusters, in a speedy way
○ Mobile Cloaking
* Designed an app which would be installed on mobiles and tablets in various
locations, to see how a URL renders on them.
* The difference in rendering on different devices would detect cloaking
* This is helpful to detect Mobile Cloaking which is ip, carrier and os based, without the need of proxy switchers
* Mentored an intern to develop the app and the server.
○ Un-reviewed Ads
* Periodically generate a dump of all the un-reviewed ads in the system
* This opened the scope for bulk actions, and different clustering mechanisms which could be applied to these un-reviewed ads and resulted in a lot of automation
* Got "Gold Award" for this project for high impact on the manual ad review system
○ Ads Review System Logs
* Generate a distributed dump of all the logs that are generated in the Ads Review
System
* This is used heavily for metric tracking and reporting and also used as a data
source to get labeled data to create machine learning models
○ Got 3 peer bonus and 2 Kudos from fellow employees, for my work.
○ Got a spot bonus for mentoring an intern
○ I was a regular interviewer
○ Similar text Detector
* Google Ads Review System is able to detect normal words, but spamers sometimes
use special characters to trick the review system. Eg AK47, @K47, where @K47
may be missed by review system and their ML models.
* This project enabled to find the closer valid words, using the similar text detection
infrastructure used at Google.
* Now the bulk review tools can capture these bad words and review these ads
○ Mobile Malware Solutions
* Platform which takes URL as input, detect and download the apk files in it and extract features from them.
* These features are then fed to ML models which detects malware in these binaries.
○ Image Ads Clustering
* Continuous pipeline which takes bunch of images as input, extract features from them, and then group the images with similar features together
* Used existing image feature extractors at Google.
* This enabled clustering and bulk reviews of the image ads.
○ Un-reviewed Ads Periodic Pipeline
* Created a continuous pipeline that runs over the un-reviewed ads, and generates
clusters from them using different types of clustering mechanisms.
* New clustering methods can be easily hooked into this infrastructure
* These clusters can now be reviewed using bulk review tools
○ Offline Machine Learning Infrastructure (OLIO)
* A generic infrastructure to create Machine Learning Models for any types of features and labeled data.
* Got First Prize for OLIO in Poster Presentation in Google Anti Abuse Summit 2014, where more than 50 teams participated.
* Used existing command line machine learning tools and created a generic
infrastructure using it
○ AdsDedup
* Project to identify duplicates in un-reviewed ads and predict the review decisions for them, based on the same ads seen previously
* These predictions act as proxy to review the clusters, in a speedy way
○ Mobile Cloaking
* Designed an app which would be installed on mobiles and tablets in various
locations, to see how a URL renders on them.
* The difference in rendering on different devices would detect cloaking
* This is helpful to detect Mobile Cloaking which is ip, carrier and os based, without the need of proxy switchers
* Mentored an intern to develop the app and the server.
○ Un-reviewed Ads
* Periodically generate a dump of all the un-reviewed ads in the system
* This opened the scope for bulk actions, and different clustering mechanisms which could be applied to these un-reviewed ads and resulted in a lot of automation
* Got "Gold Award" for this project for high impact on the manual ad review system
○ Ads Review System Logs
* Generate a distributed dump of all the logs that are generated in the Ads Review
System
* This is used heavily for metric tracking and reporting and also used as a data
source to get labeled data to create machine learning models
○ Got 3 peer bonus and 2 Kudos from fellow employees, for my work.
○ Got a spot bonus for mentoring an intern
○ I was a regular interviewer
Machine learning, Automation, Rendering, Mentoring, It, Poster, Infrastructure, Server, App, Feature, ADS, Google, Bonus, Internal, Social media ads
2011 - 2011
internship
Software Developer Intern
Google Hyderabad.
○ Groups Email migration API to archive read mail from Lotus, Microsoft exchange etc to google groups. Mentor: Ankur Jain
● Master's Research project:
○ Secure Reuse Of Disk Blocks in Minix 3.2, based on Information Flow Control
○ Implemented the Information flow Model in MINIX 3 using HiStar like labels which form a Lattice.
○ Proved that the Lattices of other labeling systems implemented in other OS like Asbestos
and Flume are either same or offer lesser granularity.
○ Used these labels attached with files to obtain improvements in latency for disk block
allocation
○ Increased security of user data by not allowing the insecure and unmonitored flow of information through disk blocks.
● Teaching Assistant for OS course at IISc
● Master's Research project:
○ Secure Reuse Of Disk Blocks in Minix 3.2, based on Information Flow Control
○ Implemented the Information flow Model in MINIX 3 using HiStar like labels which form a Lattice.
○ Proved that the Lattices of other labeling systems implemented in other OS like Asbestos
and Flume are either same or offer lesser granularity.
○ Used these labels attached with files to obtain improvements in latency for disk block
allocation
○ Increased security of user data by not allowing the insecure and unmonitored flow of information through disk blocks.
● Teaching Assistant for OS course at IISc
Research, API, Teaching, Exchange, Mentor, Developer, Security, Software, Google, Internal
My education
2010
-
2012
Indian Institute of Science
N/a, M.E
N/a, M.E
2006
-
2010
Visvesvaraya National Institute Of technology
Bachelors, Computer Science & Eng
Bachelors, Computer Science & Eng
?
-
2010
State Rank
N/a, N/a
N/a, N/a
Piyush's reviews
Piyush has not received any reviews on Worksome.
Contact Piyush Yawalkar
Worksome removes the expensive intermediaries and gives you direct contact with relevant talent.
Create a login and get the opportunity to write to Piyush directly in Worksome.
are ready to help you
and get specific bids from skilled talent in Denmark