GPUs
Inference in large Bayesian networks
Genetic linkage analysis

CV
Publications

I am working on various aspects of High Performance Computing, ranging from large-scale opportunistic grids to GPUs.

My goal is to accelerate genetic linkage analysis used to locate genomic locations of disease-provoking mutations in large genealogies. This problem is computationally intensive and may require years of CPU time. Under the hood it is mapped to the inference in very large Bayesian networks. 

The main project is Superlink-online - genetic linkage analysis system which exposes the power of tens of thousands non-dedicated CPUs in multiple grids ( Technion, EGEE, OSG, UW Madison Condor pool and Superlink@technion community grid) to geneticists worldwide.

Superlink-online uses GridBot, the policy-based system for running Bags of independent Tasks on multiple grids. My SC09 paper on GridBot is here.

I also work on accelerating genetic linkage analysis by using GPUs, and more generally, on  acceleration of  Sum-Product computations. This type of computation is used in many domains, in particular,  for the inference in Bayesian networks. My ISC08 paper is here.

Projects 

• Superlink-online: genetic linkage analysis system
• GridBot: running Bags of Tasks on grids
• Superlink@Technion: community computing grid for genetic linkage analysis
• Sum-product computations on GPUs
• High Availability Daemon for Condor (aka HAD)
• DirectorY Network Archiver and Mover  (aka Dynamo) 

Teaching

Courses

I have been teaching Concurrent and Distributed Programming undergraduate course for quite a while. Together with Prof. Schuster we developed the syllabus. I am responsible for the tutorials, some lectures, home assignments and exams...  

Previously I was TA in Object Oriented Programming and Intro to CS.

Projects

I supervised the implementation of several Condor- High Throughput Computing enhancements and addons projects at the Distributed Systems Laboratory.  In particular, Condor High Availability Daemon project  was developed from scratch and now is deployed in many production systems.

Internships

I spent  fantastic three summer months of 2007 in UC Davis working in the group of Prof. John Owens. This was the turning point in my research on GPUs.

In 2009 I visited Prof. Matsuoka lab in Tokyo Tech working on TSUBAME - the world largest supercomputer based on GPUs. The goal was to use TSUBAME for genetic linkage analysis, and it was accomplished! The preliminary results are promising.

Past

I worked as a student researcher at IBM Haifa Research Laboratory, at the Distributed System Department, conducting research on Grid Computing Systems. I am one of the authors of Dynamo - IBM directory replicator for Grid.