Secure Data Services

Design & development of sqoop-secure data services library integration for encryption and decryption of PAN and PII data. Successfully integrated the remote and local encryption methods to the sqoop mapper classes to allow for data in-transit cryptographic operations at Visa Inc.

Hadoop Cluster Performance Imporvement

Performed a study of the CPU utilization patterns, cluster configurations and scheduling algorithms at Visa Data Platform’s Dev, Q/A and Production Clusters. Analyzed the usage patterns and detected the cause of peak utilization. Optimized the process in the YARN scheduler and validated the solution by simulation of workloads

Machine Learning approaches to task partition of the OpenCL kernels

Analyzed and implemented a classification based machine learning model to determine the best device (CPU/GPU) or combination of devices (CPU+GPU) for the OpenCL kernel execution. Stochastic predictive models or Regression models where compared against the hierarchical classification approach with Support Vector Machines (SVM).

AlCoViC: Accelerated Computer Vision on Heterogeneous Coprocessors such as Intel MIC and Nvidia GPGPUs.

Comparative study of Accelerated Computer Vision Applications on CPUs, GPUs and Intel MIC-Xeon Phi. Each (co-)processors behaved differently to the type of parallelism exploited and the data access patterns.The study showed that the MIC performed comparable to GPUs when regular operations and computation patterns where used and the GPU is efficient for irregular data access and atomic operations, as expected.

Voice Forensics

A system that would identify bodily features such as height, weight, age, sex, region of origin and various other demographic information about a miscreant from the voice evidence collected. The end objective was to build an extensive, if not comprehensive, one-of-a-kind voiceprint database to enable authorities to track criminals. The project was done under the guidance of Prof. Bhiksha Raj and Prof. Rita Singh, School of Computer Science, Carnegie Mellon University.

Static and Dynamic Video Summarization Technique

An improvised static and dynamic video summarization technique, was implemented under the guidance of Dr. Manohara Pai. M. M, Associate Director, Manipal Institute of Technology. In this project, we made use of various data mining techniques such as k-means and k-nn algorithm, along with acceleration using Nvidia GPGPUs for better performance of the video summarization.

Microsoft Consulting Services India Delivery Dashboard

This was a summer project on the design and implementation of the MCS India Delivery Dashboard for customer centric delivery tracking and management under the guidance of Microsoft Consulting Services India Lead, Ms. Divya Sampath. My role in this project was to analyze the Key Performance Indicators and metrics from multiple data sources and implement the web-enable dashboard for information management and actionable decision making.

Scavenger – A smart waste management system.

Scavenger was a product of the Team Internet during the GE Foundation Scholar Meet at John F Welch Technology Centre, Bangalore during 2013. It included an Advanced Dustbin Design, Innovative Incinerator and Wireless Central Monitoring System for efficient disposal and management of waste in India buildings. My role in this project was to design a frugal method to build a smart waste management system and design the backend required for it.