I am interested in the areas of machine learning, image search,  artificial intelligence, and data mining. More specifically, I have performed research in:

• Image Search, Content Based Image Retrieval
• Multiple Instance learning
• Semi Supervised learning
• Image object/background Extraction
• Protein Binding
• Reinforcement learning in continuous state spaces
• Data mining, including classification, and clustering

Research Synopsis

Multiple Instance Learning, Doctoral Research Assistant             2002-Present CSE Department, Washington University, St. Louis, Missouri

• Conducted research and development in Multiple Instance learning, a division of Artificial Intelligence
• Developed algorithms with applications to image retrieval (CBIR) and pharmaceutical drug discovery
• Developed GEM-DD algorithm. Results compete with the top Multiple Instance algorithms
• Developed MISSL, one of the first algorithms to combine Multiple Instance and semi-supervised learning
• Developed new image representation to significantly improve retrieval accuracy in CBIR systems
• Designed and implemented Accio! CBIR system code base, now among the top image retrieval systems in the world
• Coordinated with 2 advisors and 7 students

Computer Aided Diagnosis -  Medical Image Search,                            Doctoral Research Assistant                                                                        2005-Present
School of Medicine, Washington University, St. Louis, Missouri

• Conducted joint project between machine learning lab and Electronic Radiology Lab to develop algorithms for computer aided diagnosis.
• Analyzed and worked to develop image representations appropriate for general diagnosis on MRI images.
• Worked on project to combine a set of experts, both human and computer, to better segment disease nodules.
• Participated in weekly joint lab journal club.

Protein-Protein Binding, Doctoral Rearch Assistant                        2007-Present
CSE Department, Washington University, St. Louis, Missouri

• Developed learning algorithm for clustering protein-protein binding paths.
• Combined and extended work from molecular biology, and robotic path planning
• The path clusters aid biologists to determine biological similarities and visualize a large set of paths.
• Project serves as base for developing algorithms to simulate high dimensional protein-protein binding.

Reinforcement Learning, Undergraduate Researcher                         2001-2002
CSE Department, Washington University, St. Louis, Missouri

• Performed research and development in Reinforcement Learning, a division of Artificial Intelligence, with specific focus on convergent value function approximation in continuous state spaces
• Conducted applied research enabling robots to learn in complex state spaces, such as real world environment, using feasible computational power and memory footprint
• Explored function approximation techniques for interpolatation of discrete points in a continuous state space
• Developed and tested heuristic methodology for safe interpolation of data points in reinforcement learning
• Coordinated with research advisor

MPEG-2 Compression for Internet-2, Research Intern                    2000-2001
CS Department, Kent State University, Kent, Ohio

• Served as member of project team developing advanced video streaming algorithms and their corresponding applications for active networks on the Internet-2 as part of a greater project for DAARPA
• Rewrote existing MPEG decoder in C++
• Wrote client-side interface: Windows 9x/NT environment, C++ Windows API programming
• Redesigned and implemented client-side for WinCE PDA
• Designed and implemented wireless video streaming for WinCE PDA
• Coordinated with project team to determine system needs, interfaces, and possible hardware implementation

Project Aria: Mars Rover, Vision Manager; Electronics Manager       2001-2002
CSE, Washington University, St. Louis, Missouri

• Served as part of student engineering team at Washington University working with the deputy science PI for NASA’s Mars 2003 Rover program to build a full-scale working prototype.
• Developed 3-dimensional vision system for the rover
• Formulated and coded vision algorithms
• Coordinated with team of 20 students, 3 advisors
• Lead sub-group of 3 students