Manish Kumar

• Integration with recharge/bill payment service providers: Increased success percentage to 90%.
• Dynamic switching among service providers based on different strategies (e.g., success percentage).
• Suggestion engine for numbers based on prefix, including MNP (Mobile Number Portability) database: Improved success percentage.
• Transactional and financial reconciliation: Ensured foolproof system.
• Coupon redemption framework: Reduced redundant redemption attempts by same user, resulting in higher user acquisition.
• Offers framework for different merchants using MobiKwik wallet.
• Dashboard for analytics and forecast on payments, recharges, merchant transactions, coupon redemptions, and campaign impact.

Position: Software Developer (Consultant) - June 2012 to October 2013

• Context Analyzer Engine: Developed a contextual engine for Komli's advertising platform, accurately classifying main concepts on webpages. Tools/Technologies: Java, JSON, WEKA, MySQL, Lucene, Machine learning.
• Lead Management System: Created an advertiser dashboard for aggregating leads from multiple campaigns, driving $80k in quarterly revenues. Tools/Technologies: Java, MySQL.
• Campaign Tracker (Pixel Tracker): Developed a pixel tracking API for monitoring offline campaign metrics. Tools/Technologies: Java, JSON, JavaScript, MySQL, Lucene.

Position: Software Developer (on payroll of Symphony Services Pvt. Ltd.) - Feb 2011 to May 2012

• ViZisense: Contributed to India's leading audience measurement and analytics platform, promoting transparency and higher ad spending on the internet. Tools/Technologies: Java, Concurrent programming, Shell script, MySQL, Hadoop.
• Campaign Tracker (Download Tracker): Designed and developed a mobile app distribution product, overseeing 30+ Android App Install campaigns in developed markets. Tools/Technologies: Java, MySQL.

• Log Analyzer: Developed a server traffic analysis tool that allows analysis without sharing data with third parties. Technologies used: Java, JSON, DB2, Lucene, Shell script, Perl, Servlets, XML.
• SSH session tracker: Created a resource management and human error diagnosis tool for SSH sessions. Technologies used: Perl, Shell script, Lucene, Apache.
• File processing utility: Implemented a file processing utility capable of handling various file types using XML configuration, eliminating the need for additional code. Technologies used: Perl, Shell script, XML.

• A Word Sense Disambiguation (WSD) system was developed for the fully automatic indicative English to Hindi Machine Translation System MaTra in the KBCS division. The existing mapping between English and Hindi words relied on common usage patterns, intuition, and experience, necessitating the need for a more systematic approach. The project involved exploring the field of WSD and implementing a system using supervised and semi-supervised learning techniques in machine learning. To analyze the context of target words, input sentences were converted from XML to ARFF format, using XSLT and XQuery. The transformed data was then utilized for machine learning through WEKA. The aim was to replace the existing system, which relied on common usage patterns, intuition, and experience, with a more reliable machine learning system. The tools and technologies employed in this project included Java, machine learning techniques, XML, WEKA, POS Tagger, JAWS, and WordNet.