May 2019 – Mar 2020 · 10 mos

• Real Time Fraud Modeling (playing a dual role of ML Engineer and DataScientist)
• (Scala Spark, Sagemaker, Python, Java, Kafka Streams, Boosting Models)
• Learning to Rank (pointwise classification, listwise optimization) approaches
• Pricing Initiatives

Jun 2018 – Apr 2019 · 10 mos

• Fraud Modeling (Experimented with online learning, boosting approaches)
• Recommendation Systems (lead Data Scientist)
• Initial Pricing (lead Data Scientist)

Mar 2016 – Jun 2018 · 2 yrs 3 mos

• I am part of Initial Pricing team, Search Engine Marketing team, Customer Acquisition and Analytics team
• On the Data Science front I deal with Pricing and Demand models, Sales Forecasting, User level modeling
• On the Statistical methods/Machine Learning front I deal with Embedding based neural networks, Log Linear models, Contextual Bandit models (Vowpal Wabbit), Probabilistic methods
• On the Engineering front I deal with Spark, AWS EMR, AWS Lambda, Docker, Storm, Kafka, Elasticsearch, Kibana, Athena, Teradata, BigQuery, Terraform technologies
• Collaborated with peers and mentored juniors on problems pertaining to ​Dynamic Pricing, Customer LifeTime Value, Incrementality Testing(Causal Impact), Identifying Fans, Attribution
• I lead the company wide DataScience JournalClub sessions (monthly discussions on research papers) at Ticketmaster
• [Ticketmaster Hackathon]
• Built Initial Pricing solutions for upcoming Onsales in 2016, 2017(runner-up) Hackathons
• Analyzed Fan/Bot sessions queue sorting impact on conversion rates using Onsale Simulator (2018)
• Tinkered with adding live events business context within double click search keyword bidding system (2018)
• [Talks/Conferences]
• I was a speaker at Strata Hadoop 2017 held at San Jose
• Details : https://conferences.oreilly.com/strata/strata-ca/public/schedule/detail/55937
• Attended NeurIPS (2017-2018) Conference and MIT Sloan Sports Analytics Conference (2018)

Sep 2015 – Dec 2015 · 3 mos · Los Angeles, California

• Researched end to end twitter bot detection, orchestrated bot campaign detection methods. Opted for semi-supervised learning approaches.
• Details : https://sites.google.com/site/2015analysis/
• Supervisor : Dr. Emilio Ferrara
• [Project funded by ONR]

May 2015 – Aug 2015 · 3 mos · Palo Alto, California

• I was involved in full stack development of Test Time Reduction Tool. I built the tool from scratch. Dealt with manufacturing analytics space in general. Manufactured products that go through all testing lines result in reduced production throughput rate. Test Time Reduction Tool provides the following functionality and reduces the test time of manufactured products by figuring out the important tests that are essential
• Provides a simple UI where the user can upload manufacturing data
• Tool understands the data by posing a set of questions to the user at the front end
• Generates valid configuration for backend server based on clients preferences
• Backend server runs different predictive models and sends back the result to client in a format that is suitable to the test time scenario
• Backend provides the options to run the predictive models in serial, multi-threaded, multi-processing mode based on the server capability
• Integrated different R models into the python backend stack using rpy2 package
• Technologies Used : Python, R, AngularJS, NodeJS, Javascript, CSS, Html
• Supervisor : Dr. Rumi Ghosh

Apr 2015 – May 2015 · 1 mo

• Dealt with Co-Evolutionary modeling of pay day lenders and regulators.
• Technologies Used : Berkeley Madonna differential equation solver
• Supervisor : Professor Daria Roithmayr

Jan 2015 – May 2015 · 4 mos

• Simulated vehicle routing of large scale transportation networks for Los Angeles road network
• Relative flow values at each intersection of the network are estimated through regression
• Generated quantitative metrics such as delay, throughput parameters of the network, which assist in choosing better algorithms that lead to less congestion in the network for real time traffic flows
• Experimented with Proportional queue policy and Backpressure routing policy.
• Technologies Used : C++, COM Interface, Matlab, PTV VISSIM, Visual Studio, Git
• Supervisor : Dr. Ketan Savla
• [Project funded by NSF]

Nov 2013 – Jul 2014 · 8 mos · Greater Hyderabad Area

• Researched different approaches for improving the discriminative power of mined visual parts, as part of Scene Classification problem using Concave Convex programming approach under the joint supervision of Dr. M. Pawan Kumar (Ecole Centrale Paris, INRIA) and Dr. C.V Jawahar (IIIT-H)
• Details : https://sites.google.com/site/mahesh2014research/home/formulation-doubts

May 2012 – May 2013 · 1 yr · Mumbai

• Application Developer for real time Stock/Custom-Basket/ETF/Index Pricing Engine
• Developed Monitor Server/Client which assists in monitoring status of all apps
• Developed Nightly Build Suite as part of CRD(Central Risk Desk) team
• Integrated 3 modes of transport (TCP, EMS, LBM) components into the pricing engine using which stock prices are down-streamed
• Have been part of NAM migration of Pricing Engine, APAC L3 Dev Support
• Technologies Used : C++, Boost, Kdb/q, Bash, Python, 29West

Feb 2011 – Mar 2011 · 1 mo · Greater Hyderabad Area

• Automated data analysis using Decision Trees
• Technologies Used : PMML models, Java, Jcharts, Apache Tomcat.