Mohan Krishnamoorthy

Ph.D. Reseach Synopsis & Research Statement

Projects · Experise · Accomplishments

Projects

• Stochastic optimization algorithms based on deterministic approximations (GMU) : 2014-Present
  • Purpose: Stochastic optimization algorithms that make use of the mathematical structure of the original problem are inefficient especially for real-world processes composed of complex process networks because they extract the mathematical structure using samples from a black-box simulation. The goal here is to improve the computation complexity and convergence of these algorithms for probabilistic models.
  • Contribution: Extracted the mathematical structure of the problem from a white-box simulation code analysis as part of a heuristic algorithm based on deterministic approximations to find the most optimal decision points for the system using statistics of the simulated probabilistic model.
  • Results: Experimental study on a 22-variable and 21-constraint real-world use case demonstrated that this approach significantly outperforms popular simulation-based optimization approaches.
  • See 1, 9 publications.
• Framework for composition, analysis, and optimization of real-world processes (NIST) : 2016-2017
  • Purpose: To build a system of reusable process models in manufacturing such that it is easy to use, simple, and cost-effective so that the end user can perform multiple analysis and optimization operations on these models.
  • Contribution: Built a software framework and prototype using Generic Model Environment and cloud architectures that allowed hierarchical composition, visualization, and analysis of manufacturing systems consisting of real-world processes from a reusable model repository.
  • Results: Demonstrated the prototype system to compose an hierarchical model for a real-world supply chain use case and performed simulation, prediction, optimization, and trade-off analysis using Pareto optimal graphs on this model.
  • See 2, 3 in publications.
• Reusable repository of process performance models (GMU, NIST) : 2015-2017
  • Purpose: To build a reusable repository of models for manufacturing so that analysis and optimization solutions need not be implemented de novo because it leads to cost and time intensive development of models and algorithms, which are difficult to modify, extend, and reuse.
  • Contribution: Designed and developed a reusable repository of mathematical models called performance models for manufacturing end-users with the goal of ease of use and reusability to compose and perform analysis and optimization on complex real-world hierarchical processes.
  • Results: This repository was used as the basis for a competition to crowdsource Reusable Abstractions of Manufacturing Processes (RAMP). For this competition, I also demonstrated the structure of a process performance model in an instructional webinar.
  • See 4-8, 10 in publications.
• Scientific algorithm for Primer design (LANL) : 2011
  • Purpose: To build a tool for primer design, which is difficult to do for highly variable DNA sequences and for which experimental success requires attention to many interacting constraints.
  • Contribution: Designed and developed scalable scientific algorithm for primer design that included recursive generation of combinatorial bio-barcodes of specified length with design constraints and dimer risk filtration among the generated primer constructs in C and Perl.
  • Results: Primer design tool (v1.0) was included among the HIV analysis tools (see current tool at v2.0).
  • See 11 in publications.
• Redesign of computing architecture to improve efficiency of scientific analysis (LANL) : 2010-2011
  • Purpose: To redesign scientific tools in order to ensure high performance and minimize compute time and file system usage.
  • Contribution: Designed and deployed five scientific tools using a Model-View-Controller (MVC) framework and web services using XML-RPC on the MVC model provided by Perl Catalyst and object oriented Moose libraries.
  • Results: The five tools were successfully deployed with an improvement of 35% in performance and 50% in file-system usage.
• Masters Thesis: Compression and caching in distributed file system (RIT) : 2009-2010
  • Purpose: To perform research on compression and caching algorithms to improve data fetch time in a distributed system.
  • Contribution & Results: Implemented a distributed system using the Java NIO framework and reduced data and file fetch time by 14%.
• Tree Viewer, Pruner, and Decorator (LANL) : 2009-2010
  • Purpose: To build tools that automate the selection and annotation of influenza genetic data by making the correct trade-off between speed and simplicity on the one hand and control over quality and contents of the data set on the other.
  • Contribution: Designed and developed the tree pruner and decorator tools to perform this selection and annotation for Influenza Sequence Database (ISD). This project was based on the open source project Archaeopteryx using Java Applets, AJAX, and REST web services with the JSON and phyloXML data formats.
  • Results: Pruner and Decorator tools were made available among influenza analysis tools and were also made open source.
  • See 12 in publications..
• Database and Web architecture development (LANL) : 2008-2010
  • Purpose: Design a schema to accommodate millions of records in ISD and develop a website over ISD to serve influenza analysis tools.
  • Contribution & Results: Designed a new schema in PostgreSQL. Also, developed a website over ISD using Perl, Mason Perl, HTML, XML, Java Scripts and SQL. Further, redesign of the website using jQuery and AJAX request objects yielded 23% better performance.

Expertise

• Programming Languages & Libraries: Java (expert), C (proficient), C++ (proficient), LaTex (proficient), SQL (proficient), NoSQL (proficient), Ruby (familiar), Perl (proficient), Python (competent), Shell script (familiar), R (proficient), JavaScript (familiar), jQuery (familiar), XQuery (competent), JSONiq (proficient).
• Technical Skills: Data analytics, Analytical Modeling, Data Science, Algorithm Design, Decision Optimization, Operations Research, Decision Systems, Model Simulation & Prediction, Database Management Systems, Software Development Life Cycle.
• Operating Systems: Linux, Windows 7/8/10, MAC OS/X.
• Version control: GIT, Repo, SVN.
• Functional abilities: Software Architecture, Object Oriented Programming, Distributed Business and Scientific Applications, Software Development and Testing, Data Mining and Analytics.
• Tools: Docker, Eclipse, Emacs and VI editors, Oxygen XML editor, Rational Rose, Microsoft Visio, Microsoft Office, Microsoft Visual C++, Microsoft Visual Studio, gedit, Atom.
• Mathematical modeling & Optimization Solvers: OPL, AMPL, CPLEX, Gurobi, MINOS, SNOPT, LGO, Coin OR, BARON, BONMIN.
• Internet Technology: Amazon AWS, Azure, Hadoop, OpenStack, Apache Spark.

Accomplishments