Experience

Duke Robotics Club

• United States
• Research Services
• 1 - 100 Employee

• Computer Vision Team Lead

  • Jun 2021 - Present

  • Integrating a simultaneous localization and mapping (SLAM) framework in ROS with stereo vision to automatically situate the underwater autonomous vehicle in its environment for better path planning and navigation• Exploring sonar as a method for 3d localization of objects• Placed 1st for technical document and 11th in performance in the RoboSub 2022 competition, which was the best in Duke Robotics history• Developed stereo vision using OpenCV to extract depth measurements of predicted objects from two front-mounted cameras• Mentored 20+ new club members, teaching them the basics of ROS, Git, computer vision, and deep learning

• Computer Vision Team Member

  • Aug 2019 - May 2021

  • Developed stereo vision using MATLAB and OpenCV to extract depth measurements of predicted objects from two front-mounted cameras• Assisted with the construction and training of a PyTorch object detection model to work on video feed, helped integrate the object detection model with the robot through ROS, and preprocessed and labeled 500+ images for supervised learning



Realtime Robotics, Inc.

• United States
• Automation Machinery Manufacturing
• 1 - 100 Employee

• Perception Intern

  • May 2022 - Aug 2022

  • Designed and built an augmented reality app to specify the locations for packages for a robotic palletizing application using C# and Unity Engine to increase the speed for creating palletizing configuration files • Constructed an algorithm to build 3D meshes and automatically find the dimensions of packages using a C++ Unity plugin • Created a robot tic-tac-toe demo using Python and threading with two robotic arms which could be controlled by either an AI or a human • Designed and built an augmented reality app to specify the locations for packages for a robotic palletizing application using C# and Unity Engine to increase the speed for creating palletizing configuration files • Constructed an algorithm to build 3D meshes and automatically find the dimensions of packages using a C++ Unity plugin • Created a robot tic-tac-toe demo using Python and threading with two robotic arms which could be controlled by either an AI or a human



The Johns Hopkins University Applied Physics Laboratory

• United States
• Defense and Space Manufacturing
• 700 & Above Employee

• Artificial Intelligence Intern

  • May 2021 - Mar 2022

  • Constructed a computer vision model using PyTorch to estimate the attributes of 300+ lettuce plants from top-down images • Experimented with and compared 15 different computer vision models, including using unsupervised learning with SimCLR, supervised learning with a Resnet50 model, incorporating pixel segmentation, and pretraining on a plant classification dataset • Independently created a lettuce plant segmentation algorithm using Python and OpenCV that utilized super pixels and hue thresholding to function as an image preprocessing step of the computer vision algorithm • Programmed Python scripts for pretraining deep learning models on a 300,000-image plant classification dataset • Assisted with the development of a genetic algorithm to select climate and growing setpoints to optimize the net income of a greenhouse and integrated expert knowledge from biologists into the algorithm • Built and trained a neural network to predict net income given the climate and growing setpoints for a greenhouse • Our team was given a Bravo Award for its exemplary work • Completed data selection, preprocessing, and collection for more than 30 construction sites and from 3 different satellite sources for a project to track anthropogenic changes in remote sensing imagery



Duke University

• United States
• Higher Education
• 700 & Above Employee

• SAGE Peer Facilitator

  • Sep 2021 - Dec 2021

  • I lead weekly peer facilitation sessions for a Physics course on electricity and magnetism • Involves reviewing course material, facilitating discussion about problems and concepts, and discussing problem solving and learning strategies to succeed in the course • I lead weekly peer facilitation sessions for a Physics course on electricity and magnetism • Involves reviewing course material, facilitating discussion about problems and concepts, and discussing problem solving and learning strategies to succeed in the course



Duke University Motorsports

• United States
• Motor Vehicle Manufacturing
• 1 - 100 Employee

• Electrics Team Member

  • Aug 2020 - Sep 2021

  • Designed a printed circuit board (PCB) to turn off the vehicle in case of issues with sensor signals or power supply • Created a PCB layout using Autodesk Eagle, specifying wiring paths and device locations for ideal space usage • Modeled custom device footprints and symbols in Autodesk Eagle using information from device datasheets • Researched and selected integrated circuits using information in datasheets to place in our PCB design • Designed a printed circuit board (PCB) to turn off the vehicle in case of issues with sensor signals or power supply • Created a PCB layout using Autodesk Eagle, specifying wiring paths and device locations for ideal space usage • Modeled custom device footprints and symbols in Autodesk Eagle using information from device datasheets • Researched and selected integrated circuits using information in datasheets to place in our PCB design



Duke University Energy Initiative

• United States
• Higher Education
• 1 - 100 Employee

• Duke Bass Connections: Deep Learning for Rare Energy Infrastructure in Satellite Imagery

  • Aug 2020 - Apr 2021

  • Completed further experiments that showed significant improvements in our object detection model when training with synthetic imagery• Co-authored and published a research paper in the International Geoscience and Remote Sensing Symposium (IGARRS 2021) conference and built a website to share our new results• Applying our process to transmission towers to show the generalizability of our synthetic data generation methodThis project was lead by Dr. Kyle Bradbury and Dr. Jordan Malof and was supported by the Duke Energy Initiative.

• Duke Data+: Deep Learning for Rare Energy Infrastructures in Satellite Imagery

  • May 2020 - Aug 2020

  • Trained a YoloV3 object detection model to locate energy infrastructure such as wind turbines from overhead imagery as a method to better understand pathways for electrification• Generated synthetic images and labels with CityEngine and Python scripts to supplement our training dataset and significantly improve the performance of the model, particularly when deploying the model on regions that lack real data• Built a website to share our findings, and presented these results to Duke students and facultyThis project was lead by Dr. Kyle Bradbury and Dr. Jordan Malof and was supported by the Duke Energy Initiative.



Duke Applied Machine Learning Group

• United States
• Software Development
• 1 - 100 Employee

• Radio Communication Heatmap Lead Data Scientist

  • Sep 2020 - Feb 2021

  • Constructed an android application backend using Python and elevation APIs to identify locations for soldiers to avoid radio jamming based on elevation data • Coordinated meetings, workflow, and task distribution for the team to be efficient, and guided less experienced team members • Composed and submitted a SIBR grant to apply for funds and resources through the Department of Defense • Constructed an android application backend using Python and elevation APIs to identify locations for soldiers to avoid radio jamming based on elevation data • Coordinated meetings, workflow, and task distribution for the team to be efficient, and guided less experienced team members • Composed and submitted a SIBR grant to apply for funds and resources through the Department of Defense



Duke University

• United States
• Higher Education
• 700 & Above Employee

• Medical Tool Design for Duke Hospital Surgeon

  • Aug 2019 - Jan 2020

  • Created a proof-of-concept prototype for a surgical tool designed to be used for endoscopic endonasal surgery • Went from ideation to prototyping, using the engineering design process to evaluate potential solutions and to pick an optimal solution for our problem • Created a proof-of-concept prototype for a surgical tool designed to be used for endoscopic endonasal surgery • Went from ideation to prototyping, using the engineering design process to evaluate potential solutions and to pick an optimal solution for our problem



Snohomish County, WA

• United States
• Government Administration
• 500 - 600 Employee

• Intern

  • May 2017 - Aug 2018