Jul 2019 – Present · 6 yrs 11 mos · Greater Hyderabad Area

Mar 2018 – Jul 2019 · 1 yr 4 mos · bengalore

• Responsible for building and maintaining next generation pharmacy (NGP) audit data pipelines.
• Next Generation Pharmacy (NGP) Audit Data Archiving
• Environment: Java, Spark, Hadoop, Hive and Azure
• Project Description: Built spark batch jobs to move NGP system and application audit data from various locations in azure cloud (blob storage, SQL server and etc.) to HDFS and then downstream systems. The audit data in HDFS useful for adhoc querying (hive) and data retention.

Mar 2016 – Mar 2018 · 2 yrs · Bengaluru Area, India

• Building applift dynamic retargeting product from scratch. This activity includes developing and maintaining product feed, audience data management and ad-serving systems.
• Audience Data Management
• Environment: GO, Kinesis Firehose and Redshift,
• Project Description: Created audience data management system. It includes data partners integration, saving the user in-app activity data and user segmentation. Integrated with third party data partners like tune, adjust, appsflyer, mparticle and etc. Store the user in-app activity received from data partners through API’s into database. Provided the flexibility to segment the users using user in-app activity.
• Product Feed Crawling and Caching
• Environment: Java, Aerospike, S3 and SQS
• Project Description: Download parse advertiser specific product feed and then generate.

Aug 2015 – Mar 2016 · 7 mos

• Currently i am working in Real Time Bidding (RTB) team. I worked on below projects
• 1> Integration with ad exchanges like Facebook, Website Custom Audience (WCA), Rubicon, Google and etc..
• 2> Caching techniques (memcache, redis, aerospike, bloom filter and etc.) to access data fastly.
• 3> Real time data propagation of bid logs using flume and kafka.
• 4> RTB user data cache optimization.
• 5> Intelligence bidding using machine learning algorithms like logistic regression, homing algorithms and etc..
• 6> RTB infra stack creation using AWS cloud formation.
• 7> RTB infra security using AWS security groups and nacls.
• 8> RTB metrics and infra monitoring using nagios.
• 9> Working on RTB platform stability, scalability, infra cost optimization and etc..

Oct 2014 – Dec 2014 · 2 mos

• Environment: Java, MySQL and Shell Script
• Project Description: Integrated FBX Website Custom Audience marketing channel.
• This allows us to target client Facebook ads to audiences of people who have visited
• client website/app and re-market to people who have expressed interest in client prod-
• ucts. Through this channel, we are serving 5 million impressions per day with 50%
• profit margin.

Sep 2014 – Dec 2015 · 1 yr 3 mos

• Environment: AWS, Puppet and Debian Packaging
• Project Description: Create, manage, provision and update RTB resources (machines,
• disks, auto scaling groups, ELBs, security groups, NACLs, cloud watch alarms and
• etc.) using AWS cloud formation. Centralized management of RTB machines using
• puppet. Wrote custom puppet modules to configure machines and deploy code in those
• machines. Debianize RTB packages and setup a repository with HA support.

Jan 2014 – Aug 2015 · 1 yr 7 mos

Nov 2013 – Nov 2013 · 0 mo · Bangalore

• Competition: Kaggle
• Domain: Machine Learning
• Environment: Octave
• Project Description: In this project, Neural network designed to recognize handwritten
• digits. It has 3 layers - input layer (400 pixels, 20 by 20 image), a hidden layer (300
• hidden units) and an output layer (10 labels). Training and test data is taken from the
• most popular MNIST databse of handwritten digits. It is trained on training data set
• and tested on test data set. Finally achieved an accuracy of 96% on both training and
• test data

Mar 2013 – Jul 2013 · 4 mos · Bengalore

• Environment: C, C++, Java
• Project Description: Linux on Sparc is a new platform. Right now, oracle database is not available on this. My objective is to bring the oracle database instance up on this new platform. In oracle database code, 80% is posix compliant, 18% is OS dependant(OSD), 2% is architecture specific. So, I have changed the platform (OS and architecture) specific code. Major code changes includes stack tracing, latching, javavm and makefiles. A lot of issues are resolved while building and running the database.
• The database instance is successfully ported on Linux-Sparc platform.

Aug 2012 – Jan 2014 · 1 yr 5 mos · Bangalore

• Porting database platform dependant changes from one platform to other platforms (Linux, Solaris, Hpux and Aix)
• Working on database Virtual Operating System layer. This layer deals with platform dependant services such as process management, memory management, inter process communication, disk i/o management, latching and error handling

Jun 2011 – Aug 2012 · 1 yr 2 mos · Guwahati

• Collaborative prediction refers to predicting the users rating on each unrated item based on the users previous ratings on items. In this work, we proposed a method to collaborative prediction problem which is based on the matrix factorization and uses the ratings alone for prediction. In this thesis work, we are trying to propose a method which increases the prediction accuracy than previous methods. This method is based on matrix factorization. It uses the both structural information as well as node specific information

Jan 2011 – Apr 2011 · 3 mos · guwahati

• Role: Designer and Core Developer
• Environment: C
• Project Description: In this project, I implemented a crawler that crawls the web pages, a Patricia trie to index the terms in the documents, searching the queries and displaying the resultant documents using TF-IDF and BM25 Ranking mechanisms.

Oct 2010 – Nov 2010 · 1 mo · guwahati

• Role: Designer and Core Developer
• Environment: PHP, Oracle, Apache Tomcat Server
• Project Description:In this project, we created and maintained the details of Salesman, Vendors and Items supplied to the Mini Market. All the day-to-day transactions are maintained as transaction data and affected in Mainstock. Mini Market has Billing and Report sections. In Billing, each customer is billed at the salesman desk. Reports can be generated on daily, weekly and monthly basis according to counter wise, salesman wise, item wise and stock availability.

Jun 2010 – Aug 2012 · 2 yrs 2 mos · Dispur, Assam, India

• Master of Technology

May 2009 – Oct 2010 · 1 yr 5 mos · Bengaluru, Karnataka, India · On-site

Dec 2008 – May 2009 · 5 mos · hyderabad

• Role: Designer and Core Developer
• Environment: .Net Framework
• Project Description: In this project, we implemented Image Steganography in which a text le is embedded into an image in such a way that the changes in the image are
• unnoticeable and sent to the receiver. At receiver side, a password is used for demerging the text le from the image.