Top Scribbles
About Education Work Experience Skills Publications Projects Resume Contact

Kiran Raj Samarthyam

Agnostic, knowledge seeker, story teller, ideator, coder, wanna be entrepreneur!

About me

Hello World! I am a research student currently pursuing M.S by Research at IIIT Hyderabad under the guidance of Dr. Kishore Kothapalli (CSTAR). My areas of interest include High Performance Computing (HPC) & Human Computer Interaction (HCI).


M.S by Research in Computer Science, CGPA - 9.16
2013 - Current, IIIT-H, Hyderabad, India.
B.E in Computer Science Engineering, Percentage - 82%
2006 - 2010, St.Joseph's College of Engineering, Chennai, India.

Work Experience

Research Assistant, Web Application Developer
July 2014 - Present, IIIT Hyderabad.
Teaching Assistant, Data Structures
Jan 2014 - May 2014, IIIT Hyderabad.
Web Application Developer
2012 - 2013, Freelancer; Co-founded 2 failed start-ups Know-N-Share, Offr.
Programmer Analyst
2010 - 2011, Cognizant Technology Solutions, India.


C, Python
Parallel Programming
Open MP, CUDA, OpenACC (Thrust, CUSP, CuSparse libraries)
Web Development
HTML, Javascript, CSS, jQuery, PHP
Database Management
Other Technologies/Tools
Web2Py, Laravel, CodeIgniter, Ariba 9r1, Git
Operating Systems Used
Windows, Linux (Ubuntu), OSX


[Published] Kiran Raj Ramamoorthy, Dip Sankar Banerjee, Kannan Srinathan, Kishore Kothapalli, A Novel Heterogeneous Algorithm for Multipying Scale-Free Sparse Matrices, to IEEE-IPDPS ASHES 2015 (International Workshop on Parallel and Distributed Scientific and Engineering Computing)
[On Going] Kiran Raj Ramamoorthy, Hardhik Mallipeddi, Kishore Kothapalli, Nearly Balanced Work Partitioning via Sampling for Heterogeneous Algorithms


Source to Source Translator using Rose Compiler
A source-to-source compiler translates between programming languages that operate at approximately the same level of abstraction. In this project, we have tried to develop a source-to-source compiler that automatically translates the given C program into another C program with instructions replaced for the target platform (Universal Multifunction Accelerator [UMA], developed by Manjeeera Digital Systems) that supports an extended instruction set. In order to accomplish this task we take help of Rose compiler, an open source compiler infrastructure to build source-to-source compilers.
Finding Convex Hull using Cache Aware/Software Prefetching based GPU Algorithms
In this project, we studied different parallel convex hull algorithms for GPU and their memory access patterns. We then identified the best parallel convex hull algorithm and introduced cache aware software prefetching techniques to improve efficiency. Also we conducted experimentations using latest GPUs like NVIDIA GTX 580/680 that has cache system and reported the results for comparison.
Face Recognition using Eigen Faces
We implemented a face recognition system using PCA (Principle Component Analysis) and SVM (Support Vector Machines). PCA reduces the high dimensional image to a low dimensional space (referred to as eigen space). Computing the eigen vectors of scatter matrix (of the training samples) and selecting Top k eigen vectors are two key tasks in this process. We have implemented the above techniques on standard datasets maintained by Yale University, CMU University and also on a real time dataset generated during CSE471(Statistical Methods in AI) class at IIIT Hyderabad. We also performed validation & verification experiments on these datasets with the above mentioned techniques, recorded results and inferences.
POS Tagging and CYK Parsing for Indian Languages
We implemented POS taggers for Hindi, Tamil and Telugu using supervised (Hidden Markov Model) and unsupervised (clustering) approaches. We also used CYK algorithm on datasets of these languages. Python was used for programming.
Effective Text Categorization Using Compactness & Position of First Appearance
Our project automatically sorts a set of documents into categories from a predefined set. In constrast to using the traditional term frequency values solely, including the distributional features requires only a little bit additional cost, while the categorization performance was significantly improved.




Research Email Gmail Facebook Twitter LinkedIn GitHub