Staff AI Scientist

• May 2023 - Present

AI Scientist

• Mar 2019 - May 2023

Research Assistant

• Jan 2013 - Dec 2019

Px system allows user to model phenomenon having relational structure as well as uncertainty. The semantics of such models is given by Distribution Semantics. It can be used for learning as well as probabilistic inference. Conducted research and developed the following novel algorithms and their prototype implementations. The implementations were done using a mix of XSB Prolog, C, and GNU scientific library. • Ordered Symbolic Derivation Diagrams: Use constraints to represent the… Show more Px system allows user to model phenomenon having relational structure as well as uncertainty. The semantics of such models is given by Distribution Semantics. It can be used for learning as well as probabilistic inference. Conducted research and developed the following novel algorithms and their prototype implementations. The implementations were done using a mix of XSB Prolog, C, and GNU scientific library. • Ordered Symbolic Derivation Diagrams: Use constraints to represent the explanations for an observation in a compact fashion and results in speedup over naive probabilistic inference. • Approximate Inference: Markov chain Monte-Carlo (MCMC) based inference is used when exact probability computations are intractable. • Lifted Explanation Graphs: Use constraints to represent the explanations compactly in the presence of independent identically distributed (i.i.d) random variables and provides speedup over naive probability computation in some problems. Show less Px system allows user to model phenomenon having relational structure as well as uncertainty. The semantics of such models is given by Distribution Semantics. It can be used for learning as well as probabilistic inference. Conducted research and developed the following novel algorithms and their prototype implementations. The implementations were done using a mix of XSB Prolog, C, and GNU scientific library. • Ordered Symbolic Derivation Diagrams: Use constraints to represent the… Show more Px system allows user to model phenomenon having relational structure as well as uncertainty. The semantics of such models is given by Distribution Semantics. It can be used for learning as well as probabilistic inference. Conducted research and developed the following novel algorithms and their prototype implementations. The implementations were done using a mix of XSB Prolog, C, and GNU scientific library. • Ordered Symbolic Derivation Diagrams: Use constraints to represent the explanations for an observation in a compact fashion and results in speedup over naive probabilistic inference. • Approximate Inference: Markov chain Monte-Carlo (MCMC) based inference is used when exact probability computations are intractable. • Lifted Explanation Graphs: Use constraints to represent the explanations compactly in the presence of independent identically distributed (i.i.d) random variables and provides speedup over naive probability computation in some problems. Show less

Teaching Assistant

• Jan 2011 - Jan 2013

Managed large classes (100+) as teaching assistant by conducting recitations, grading, lab sessions and office hours. UnderGrad/Grad level Courses: (i) C programming, (ii) Algorithms and Data Structures, and (iii) Network Programming. Managed large classes (100+) as teaching assistant by conducting recitations, grading, lab sessions and office hours. UnderGrad/Grad level Courses: (i) C programming, (ii) Algorithms and Data Structures, and (iii) Network Programming.

Software Engineer

• Jun 2010 - Jan 2011

Involved in design and development of a stress test framework (“Quantum In- strumentation”), which provides test coverage for Teradata RDBMS product. Developed framework to auto-generate simulated asynchronous failure events (asynchronous aborts and query rewrites) for increasing the overall test coverage. The software was built using Java, Perl and SQL. Involved in design and development of a stress test framework (“Quantum In- strumentation”), which provides test coverage for Teradata RDBMS product. Developed framework to auto-generate simulated asynchronous failure events (asynchronous aborts and query rewrites) for increasing the overall test coverage. The software was built using Java, Perl and SQL.

Associate Software Engineer

• Aug 2007 - Jan 2010

Dynamic tasking prototype of IBM POE: Involved in design and development of dynamic tasking prototype for IBM parallel operating environment (POE). Lead the effort in design of dynamic tasking feature for Message Passing Interface (MPI) 2.0 and developed resource management architecture (an iterative TCP/IP server), which arbitrates requests for resource usage by MPI processes. This software was developed using C and MPI library. SNAP genetic programming toolkit: Involved in development… Show more Dynamic tasking prototype of IBM POE: Involved in design and development of dynamic tasking prototype for IBM parallel operating environment (POE). Lead the effort in design of dynamic tasking feature for Message Passing Interface (MPI) 2.0 and developed resource management architecture (an iterative TCP/IP server), which arbitrates requests for resource usage by MPI processes. This software was developed using C and MPI library. SNAP genetic programming toolkit: Involved in development of functional module to perform scan- chain optimization using SNAP programming tool kit. Conceptually, a “multi-depot vehicle routing” problem (NP-complete), which is suitable to be solved by genetic algorithms. System developed using C++. Show less Dynamic tasking prototype of IBM POE: Involved in design and development of dynamic tasking prototype for IBM parallel operating environment (POE). Lead the effort in design of dynamic tasking feature for Message Passing Interface (MPI) 2.0 and developed resource management architecture (an iterative TCP/IP server), which arbitrates requests for resource usage by MPI processes. This software was developed using C and MPI library. SNAP genetic programming toolkit: Involved in development… Show more Dynamic tasking prototype of IBM POE: Involved in design and development of dynamic tasking prototype for IBM parallel operating environment (POE). Lead the effort in design of dynamic tasking feature for Message Passing Interface (MPI) 2.0 and developed resource management architecture (an iterative TCP/IP server), which arbitrates requests for resource usage by MPI processes. This software was developed using C and MPI library. SNAP genetic programming toolkit: Involved in development of functional module to perform scan- chain optimization using SNAP programming tool kit. Conceptually, a “multi-depot vehicle routing” problem (NP-complete), which is suitable to be solved by genetic algorithms. System developed using C++. Show less