Justin Xiang

Experience

• 

  Weill Cornell Medicine

  Mar 2021 - Jan 2023

  Scientific Researcher

  Completed two research projects developing deep learning and machine learning for better patient outcomes under the mentorship of Prof. Fei Wang.1. Improving Cell Type Classification in Histology Images using Active Learning:- Used stained histology image data of mice suffering from inflammatory arthritis and labeled 2,800 cells from six classes for single-cell classification and collected the initial set of features by employing a HoverNet segmentation model and Fiji feature extraction tools.- Generated 40x more features through statistical calculations of cells within various neighborhoods.- Used active learning sampling strategies to demonstrate a high learning efficiency with minimal amounts of data and achieved drastic improvements in accuracy2. Deep Learning-Based Pose Estimation for Sports Injury Risk Detection:- Developed machine learning models for classifying risk of injury in MLB pitching videos using optic flow frames and a deep learning-based pose estimation model.- Tested the effectiveness of the new feature in multiple classification models and performed linear combination to improve the accuracy of the models by a significant margin. Show less

• 

  CUAir

  Oct 2023 - now

  Intelligent Systems Software Engineer

  Working on automatically detecting and identifying targets from images captured by the plane while in flight. Developing and testing algorithms for path planning and obstacle avoidance. Trained and tested various large vision models for object detection and classification with high accuracy used at competition. Set up a PostgreSQL-based backend for easily collecting and managing training data over time and worked on ground to plane server communication to process images from flight.

• 

  Cornell University

  Feb 2024 - now

  Research Assistant at the Cornell SciFi Lab

  Currently conducting research on using machine learning to perform pose estimation and activity recognition with acoustic-sensing wearables. Developed and trained a clap detector to detect the exact point of hand contact from various angles of video input using deep learning to automate visual and acoustic synchronization for multiple lab projects.

• 

  Cornell Tech

  Jun 2024 - now

  Summer Research Intern

  Conducting research at the Artificial Intelligence & Robotics Lab under Professor Angelique Taylor on human-robot collaboration using a wearable armband in clinical health settings by leveraging machine learning for multimodal EMG and IMU gesture recognition and user identification. Built both a ROS1 and ROS2 architecture for robot communication and control using the armband. Developed a downloadable, full stack desktop application using the Electron.js framework for customizing and training gesture recognition models. Show less