Teaching Assistant for the Search Engines: 11-442 / 11-642 course offered by Prof. Jamie Callan• Graded Assignments, reading summaries and exam answer sheets • Focused mainly on exact match and best match models, pseudo-relevance feedback, search diversification, query expansion and learning to rank

• Part of Document Sensei Team, Adobe’s AI Platform which powers Document Intelligence• Building Machine Learning solutions in Caffe for the Adobe Scan app

Pursued research in the Machine Learning Department under Prof. Tom Mitchell as part of the "NELL: Never-Ending Language Learner" research group. Worked on two main projects :• Deep Question Answering (QA) for Personal AgentsDeals with the task of multi-relational QA over personal narrative, which combines elements of both structured QA over knowledge bases, and unstructured QA over narrative as part of Yahoo InMind Project• Zero-Shot Content Learning for Email ClassificationAim is to train models to learn natural language descriptions of classes and classify a given email into one or more classes, without relying on any labeled data

As part of the "Machine Vision" team, worked on a range of problems like Attribute Extraction, Visual Similarity and Dominant Color Extraction employing Machine Learning, Deep Learning and Natural Language Processing techniques.• Implemented an AI-Designer which generates computer designed apparels by employing Autoencoder and Adversarial Networks• Developed a Machine Curation solution in Python to predict trends and make better recommendations, by building an ensemble of a Deep Learnt Image Model and a Random Forest based Text Model• Modeled a solution to identify the graphic type on T-shirts using Unsupervised Learning and Convolutional Neural Networks

"​ Agile Desktop Software Development using Behavior Driven Approach Approach in Java "• Developed a functionality in Java to automate the processing of cash flows by employing Spring, Hibernate and Eclipse RCP frameworks and JFace and Standard Widget Toolkits• Leveraged on Scrum and behavior-driven development methodology to develop a scalable system with high throughput and uptime• Tested using Cucumber, Selenium and JUnit and enhanced productivity by eliminating manual effort

" Design and Implementation of Cloud Data Synchronization on Android Platform Optimized for Slow and Erratic Network "• Designed and implemented a novel cloud data synchronization mechanism between Android clients and Rails-based server to save battery power and ensure fault tolerance• Optimized for slow and erratic network by compressing data, batching network as well as I/O bound operations and adopting a server-side conflict resolution mechanism• Measured performance under various network conditions and the new method was found to be 36% faster and consumed 23% less bandwidth

Pursued research in the "Supercomputer Education and Research Centre" under Dr. J Lakshmi, Principal Research Scientist as part of CADlab" Measuring and Identifying Application Performance Bottlenecks on Linux Using Perf Tool "• Identified and measured application performance bottlenecks on Linux using Perf• Tested Perf on various application programs to understand and analyze the data it gathers to correlate with application execution and system behavior• Generated flame graphs from the profiling data to identify the hot paths of the program quickly and accurately and further employed this to optimize application performance