I have worked on academic projects as well as on industrial projects.
Projects at Electronic Arts
Scoobie: Adhoc Analytics Platform.
Technology: Java, Hadoop, Hive
· Developing a multi-tenant querying platform for accessing game play data using intelligent queuing to provide fairness in resource utilization amongst hundreds of users as well as enabling resource monitoring.
Lighthouse: Monitoring System.
Technology: Java, Hadoop, Hive, Oozie, Primefaces, JSF, Spring Web flow
· Developing a monitoring application for reporting SLA, health status from Ganglia from multiple Petabyte clusters. It reports current and historical usage stats to enable better capacity planning in closed-loop system.
Projects at Windows Azure
Analytics Platform for Azure Fabric Fundamentals.
Technology: C#.Net, SQL Azure, Windows Azure
· Developing a Cloud based framework for accessing deployment statistics, modules for accessing different event to analyze fundamental aspects (availability, performance) of Windows Azure.
RedBox: Running Fabric (Distributed Kernel of Windows Azure) on a box
Technology: C#.Net, Managed C++, Windows Azure
· Develop a framework with inventory to simulate multiple nodes on a desktop.
· Develop simulated (load-balancer, power etc.) drivers to run latest version of kernel.
Design Cloud-based Search Engine for Windows Azure logs.
Technology: C#.Net, SQL Azure, Windows Azure
· Developing a distributed search engine for Windows Azure production logs. The indexer (big and small), merger are distributed across different cluster to process logs locally at each cluster.
Projects at Credit Suisse
Develop Grid Computing infrastructure for Quant applications.
Technology: SUN Grid Engine 6.1, MPI.
· Set up the Grid across HP Blades (running SLES9 OS) and Sun Ultrasparc T2000 processors for Quant applications. Installing MPI libraries and tested some parallel codes on the Grid.
Develop a MATLAB-MEX interface MATLAB-MEX to Access Market Data from QAI.
Technology: C++, MATLAB
Automating Mars, a RMM (Risk Measurement and Management) Tool.
Technology: C# .Net, Oracle 10g
Academic Projects
Jitter (Noise)-aware scheduling on top of the default FIFO scheduler of Hadoop.
Technology: Hadoop, Java, C++
· Identifying sources of noise (Jitter) for Hadoop and developing a scheduling policy on top of the Hadoop FIFO scheduler to mitigate the jitter-effect to improve scalability and performance.
Power-Performance optimizations for Multithreaded Numerical Algorithms on Multicore Processors.
Technology: C, C++, Simics, CACTI, dinero, POSIX, OpenMP, LAPACK-BLAS.