Jun 2017 – Nov 2019 · 2 yrs 5 mos · Bengaluru Area, India

• Developed and owned the critical revenue model and also responsible for rolling out significant features on the core revenue algorithm. Key technologies used are Python, SQL Databases, NLP, sklearn, xgboost, Scala.
• Working on development of robust modules for hybrid optimizer which offers the envelope of AdCloud
• optimization and advertiser goal optimization offered by publisher platforms.
• Worked on development of a scalable system which syncs the touchpoints and attribution data from search engine to optimally correct keyword bids using feature engineering and xgboost.
• Worked on developing a realtime API which intelligently recommends daily budgets to the client for a
• portfolio based on user input weekly budget for improving overall ROI and accuracy.
• Worked on building a scalable and reliable system to solve new issues and answer FAQs based on previously solved tickets and documentations provided about the components of the system.
• Worked on building a scalable low latency module to simulate the behaviour of search engine in a controlled environment to be used for intelligent bidding of zero impression keywords.
• Mentored intern for rolling out significant features for intraday optimization to improve the time complexity of pushing 1 million bids every hour and test using aggressive/conservative simulation of search engine.

Jun 2015 – Jun 2017 · 2 yrs · Bengaluru Area, India

• Worked on development of a hierarchical revenue model based on the historical attribution data for search engine advertising which improved performance in terms of accuracy and time complexity.
• Designed a scalable module for assisting clients to restructure their campaigns based on the creatives of the ads and static business logic inspired feature engineering to maximize the revenue.
• + Worked on building scalable regression models based on the advertiser’s cost and attribution data across search, display and email channels using Hadoop taking into account the seasonality of the data.
• + Worked on developing a module to provide budget allocation across various channels of advertising using convex hull algorithm based on the optimal allocations for portfolios across channels.
• + Mentored intern for finding a suitable mechanism to bid zero impression keywords based on the similarity with the head keywords using natural language processing and clustering.

Nov 2014 – Jun 2015 · 7 mos

• Worked on a project based on support vector machine. The work is based on studying a new SVM
• classifier based on pinball loss technique to get better results than the conventional hinge loss procedure while studying imbalanced data in noisy environments. The SVM classifier is being trained using various re-sampling techniques over the training set examples. Evolutionary algorithms have been applied to the samples to obtain better results.

May 2014 – Jul 2014 · 2 mos · Grenoble Area, France

• The work consisted in the development of a new algorithm for multi-class
• classification problems. With respect to the state of art the originality of the approach is to
• reduce the problem into a binary classification problem by considering a new cost function
• which measures the average number of classes getting higher scores than the true class of
• examples by the prediction function that is learned. Found a new representation of the pairs of
• documents and classes using neural networks.

May 2013 – Jul 2013 · 2 mos · Kharagpur Area, India

• Single instruction CPU and regular CPU design in Virtual Lab for Computer
• Architecture. Verilog code generation from gate level (cyclic and acyclic) circuits, flipflops
• and multiplexers and also did a GUI based work on Algorithm State Machine (ASM).
• Developed a GUI to facilitate the capture of a given railway yard diagram for verification
• of railway signaling systems.
• BDD based logic synthesis for countering differential power attack.