Migrate and optimize C++ software components and algorithms for NASA's OSAM-1 satellite servicing robot.● Transitioned multiple C++ components from the Robotic Demonstration and Test (RDT) / Algorithm Development Platform (ADP) software to the robot flight software (RSW), ensuring platform consistency and flight readiness.●Implemented and supported testing of an existing heater control algorithm using C++ to maintain a temperature differential of less than 1° Celsius across the force-torque sensor, optimizing its accuracy.● Participated in day-to-day iterative software development, ensuring high code quality and system compatibility through rigorous unit and integration testing.● Assisted in reviewing and revising the overall RSW startup infrastructure to ensure that it loads and starts in a flight-like manner when running on the VxWorks operating system environment.● Developed a robust launcher program in a Linux virtual machine environment that uses CPU affinity to minimize context switching and interrupts, achieving better real-time performance and facilitating local RSW development and testing cycles.

Teach graduate-level robotics courses: ● Software Development for Robotics (ENPM700 - https://mage.umd.edu/enpm700-0) ● Perception for Autonomous Robots (ENPM673 - https://mage.umd.edu/enpm673)

• Served as team lead; created reference setup, provided training sessions, and managed team's work of migrating legacy Java software to run on a private cloud platform. • Integrated open-source machine learning and analytics into a multi-tiered architecture. Analytics included speech recognition, speaker recognition, scene classification, optical character recognition, and text sentiment analysis. • Trained customized speaker and face recognition systems. • Developed ATAK (Android Team Awareness Kit) plugins and ported Linux software to run on Android OS.

1.) RESEARCH IN SATELLITE IMAGE UNDERSTANDING: • Developed a concise-set representation of multispectral satellite data to greatly reduce the manual work of ground-truth annotation needed for testing and training satellite land-cover classifiers. • Created satellite land-cover classifiers using Support Vector Machine and Bayesian Machine Learning. • Designed and implemented PIMSIR, a software tool that allows rapid visualization of overlapping satellite data covering a large geographic area. • Designed and implemented a generic road-following framework based on OpenStreetMap (OSM) and satellite images to extract objects or markings along and near the roads. • Researched and experimented with different skeletonization approaches to create vectorized road network using OSM and satellite images. • Prototyped and demonstrated an interactive, user-in-the-loop, active learning system for extracting buildings and marinas from satellite images. • Researched and applied computer vision and machine learning algorithms to extract large man-made objects such as water towers and other specific building types from high-resolution panchromatic satellite images.2.) RESEARCH IN CLOUD COMPUTING AND BIG DATA ANALYTICS: • Created a simplified and efficient distributed Map-Reduce framework based on the Blackboard interprocess communication architecture. • Applied locality-sensitive hashing and big data processing techniques to tackle the problem of clustering very large datasets. • Deployed, managed, and provisioned a cluster of high-end servers running OpenStack and Eucalyptus Cloud.3.) ROBOT MODELING AND CALIBRATION: • Modeled custom robotic arms and grippers through inverse kinematics simulation and robot calibration. • Designed and fabricated calibration artifacts needed for custom robot calibration.

1.) RESEARCH IN MEASUREMENT SCIENCE FOR 6-D POSE SYSTEMS: • Participated in the ASTM E57.02 subcommittee, which focuses on providing industry with standard test methods for evaluating 6-D pose measurement systems. • Developed test and evaluation methods for the static measurements of 6-D pose systems. • Provided ground-truth measurement support for calibration and evaluation of various commercial positioning sensors/systems.2.) VISUAL ANALYTICS TOOLS AND DATA VISUALIZATION SOFTWARE: • Designed and led the development of MagicScore, a real-time interactive scoring software used by the adjudicators during the Multi-Autonomous Ground-robotic International Challenge (MAGIC2010) ― a 7-day outdoor mapping competition sponsored by the U.S. and Australian Departments of Defense. • Created CTAviewer, a visual analytics software that analyzes, displays, and tracks multiple realtime positional data produced by an Ultra WideBand (UWB) outdoor positioning system capable of monitoring an outdoor area as large as 80,000 square meters.3.) GROUND TRUTH COLLECTION AND PERFORMANCE EVALUATION: • Created software to collect official RGBD ground-truth data for the NIST and Willow Garage Perception Challenge ― a prize computer vision competition held at the 2011 IEEE International Conference on Robotics and Automation (ICRA). • Devised a protocol using a function generator to trigger and synchronize data acquisition between a laser tracker and multiple FireWire cameras.4.) RESEARCH IN COMPUTER VISION FOR AUTONOMOUS DRIVING: • Developed a Laser Radar (LADAR)-based vision software for use on industrial automatic guided vehicles (AGV) to detect and locate palletized goods located in an unstructured and cluttered environment. • Assisted in the modeling and calibration of Laser Radar (LADAR) sensors used by the Demo III eXperimental Unmanned Vehicle (XUV) project sponsored by the U.S. Army.

RESEARCH IN COMPUTER VISION FOR AUTONOMOUS DRIVING: • Developed a LADAR-based obstacle detection algorithm that not only detects positive and negative obstacles (e.g., rocks and depressions on the ground), but also detects overhanging obstacles (e.g., tree cover).