Hi, I'm James -- a data scientist with a background in biomedical engineering and experience in both corporate startup and academic research environments. I enjoy building end-to-end AI solutions to complex problems, especially involving disease detection and diagnoses. As an engineer, I always believe that the simplest solution is the best solution. It's getting to the simplest solution that is the hard part, and only by diving into the data and gaining a deep understanding of it can you trim off all of the fat and create a truly useful tool. In my 7+ years of work, I have applied this mentality to the many AI/ML projects I have undertaken.
In 2018, I started my journey as an undergraduate research assistant at the Quantitative BioImaging Laboratory, where I built my first neural network for predicting tumor locations in hyperspectral imaging scans of the brain. I authored two publications, both involving augmented reality for surgical guidance.
In 2020, I joined the Center for Imaging and Surgical Innovation, a joint lab between UT Southwestern Medical Center and The University of Texas at Dallas. There, I was drawn further into the world of deep learning and data science. I learned how to handle large amounts of patient data efficiently and kept up to date on the latest state-of-the-art models and AI trends. Leveraging these skills, I tackled several real-world medical problems, including motion correction of the kidneys in long-duration MRI time scans and authored three research papers.
I went on to work at Croptix in 2023, an agtech startup focused on improving large-scale plant health and disease detection using novel sensor technologies and cloud/AI solutions. Having built an end-to-end cloud pipeline for real-time, in-field nutrient deficiency detection, I presently continue to develop new models and methodologies to boost performance further while using my insights of statistics and data to present any new findings to clients and stakeholders.
GPA 3.96
Successfully defended my thesis, entitled "Segmentation, Registration, and Topography-based Feature Extraction for Placental MRI"
GPA 3.92
4D MRIs of the kidneys are often acquired with long acquisition times, which can lead to motion artifacts. I built a two-stage, coarse-fine deep learning pipeline with convolutional neural networks and vision transformers to correct motion artifacts in kidney MRIs.
View Project | View Paper
I designed a 3D UNet model to segment out the placenta and uterine cavity from prenatal MRI scans with high accuracy to aid in detection of placenta accreta spectrum.
View Project | View Paper
Most of the relevant features of the placenta for diagnosis of placental complications lie in its location and shape. I created a novel method to extract 2D topographical features from 3D prenatal MRI scans to improve current automated diagnosis methods.
View Project | View Paper
I trained a neural network to segment out the tumor from hyperspectral images of agar-based brain phantoms, then I built an app for the Microsoft HoloLens to visualize the boundaries in real-time.
View Paper
I built a VR app with various fun and engaging exercises for stroke victims to regain mobility in their hands. It interfaces seamlessly with a custom Arduino-based sensor glove which measures finger flexion and strength.
View Project | View Demo | View Paper