Sravanthi Kota Venkata



Download 14.89 Kb.
Date07.08.2017
Size14.89 Kb.


Sravanthi Kota Venkata

9258 Regents Road Apt A  La Jolla, CA 92037


Phone: (858) 334 3223  E-Mail: sravipooji@gmail.com

http://cseweb.ucsd.edu/~skotavenkata

Objective


To obtain a challenging full-time position in Processor Architecture, Memory subsystem and Performance Analysis and Modeling with scope for research and development.

Research Interests


Computer Architecture - core microarchitecture design, memory hierarchy and system design. Compiler Design. Performance modeling and analysis. Applied Computer Vision.

Education


M.S. in Computer Science and Engineering

University of California, San Diego

Fall 2008 – Spring 2011 (expected). CGPA: 3.64/4.0.

B.E.(Hons). Electrical and Electronics Engineering

Birla Institute of Technology and Science, Pilani

Fall 2002 – Spring 2006. CGPA: 8.04/10.0

Relevant courses - computer architecture, adv. parallel computer architecture, low power VLSI design, adv. compiler design, operating systems, computer vision, algorithm design and analysis.

Technical Proficiencies


Languages – C, C++, Python, MATLAB, Java

Technical Skills – Parallel Programming (OpenMP, CILK), Functional Programming (Haskell), Assembly Programming (DSP, RISC, x86, MIPS), Embedded Systems and DSP

Research Experience


Graduate Student Researcher, University of California San Diego

Advisor: Prof. Michael B. Taylor Summer ‘08 – Present

Attacking Memory and On-chip Network Latencies in Multi-core Architectures [1]

The performance of multi-core systems is severely limited by the memory and on-chip communication latency. In this project, we minimize them by finding efficient mapping of memory accesses, and optimally redistributing data so as to place them closer to the execution nodes, thus reducing off-chip memory accesses.


Parallel Programming on Many-core Systems

The goal of this project was to parallelize Computer Vision applications using CILK and OpenMP programming models such that speed-up scales with the number of cores. This was very challenging considering the negative interference of shared resources and the overheads introduced by the parallel programming models. The project involved applying complex code and data transformations techniques and evaluating them on 32-core systems. This involved understanding the system-level design of 32-core machines and optimizing the software to map efficiently onto hardware.


Automating Parallelism Discovery and Planning [2, 3]

Developed compiler and runtime techniques such as dynamic critical path analysis to automatically expose parallelism and predict speedup of target programs. We applied these techniques on SD-VBS (The San Diego Vision Benchmark Suite), among others, to expose the underlying parallelism that could be extracted using multi-core architectures.


SD-VBS: The San Diego Vision Benchmark Suite [1]

Developed SD-VBS, which is a suite of diverse vision applications drawn from the vision domain. Vision applications have a fair amount of parallelism, which makes them a good candidate for formulating future multi-core and parallel architectures. SD-VBS is intended to help architects, compiler writers, and system designers to study the construction of future systems that excel at vision-oriented applications.



Work Experience


University of California, San Diego Fall 2010

CS Teaching Assistant – Graduate Computer Architecture. Facilitated a graduate class on principles and design of architecture and advanced memory. Took initiative to design new course material and lead discussions on emerging research in core and interconnect design.

Sarnoff Innovative Technologies, Bangalore July 2006 – July 2008

Research Engineer – Applied Vision. Developed DSP portable image processing and vision algorithms for real-time applications such as Adaptive Cruise Control, Pedestrian Tracking, Lane Departure Warning System. This involved porting vision applications from Matlab to C and assembly language. For this purpose, the applications had to be converted from floating point to fixed-point assembly, in addition to optimizing them for the underlying DSP pipeline architecture and memory model. Finally, these DSP-compatible vision applications were packaged with real-time performance of >25 fps.

Analog Devices, Bangalore July 2005 – Dec 2005

Intern – DSP Software. Worked on porting audio/video codecs, such as, H.263 and H.323 standards onto Blackfin DSP processor. Implemented system-C and assembly codes compatible with DSP for the audio-video codecs that perform at real-time speed of >25 fps.

Birla Institute of Technology and Science, Pilani Jan 2005 – July 2005

EE/CS Teaching Assistant – Microprocessor Design and Integration.

Publications


[1] SD-VBS: The San Diego Vision Benchmark Suite

Sravanthi Kota V, I Ahn, D Jeon, A Gupta, C Louie, S Garcia, S Belongie, M B. Taylor.



IEEE Workload Characterization Symposium, 0:55–64, 2009.

[2] Kremlin: Like gprof, but for Parallelization (Winner “Best Student Poster” award)

D. Jeon, S. Garcia, C. Louie, Sravanthi Kota Venkata, M.B. Taylor

Principles and Practice of Parallel Programming (PPoPP), February 2011

[3] Bridging Parallelization Gap: Automating Parallelism Discovery and Planning

S. Garcia, Donghwan Jeon, Chris Louie, Sravanthi Kota Venkata, Michael B Taylor.



USENIX HOTPAR, June 2010.

[4] iRain: Data Water Cycle for Raining User With Information

Ganesh Venkatesh, Sravanthi Kota Venkata, and Robert Chen



WACI VIII at ASPLOS XV, March 2010.

References will be provided on request


Download 14.89 Kb.

Share with your friends:




The database is protected by copyright ©ininet.org 2020
send message

    Main page