From brain imagingto sports simulation.
I’m Javid, an AI/ML researcher at Penn Medicine and the founder of Scoutics. I develop methods for brain image analysis and am building simulation-based tools for sports intelligence.
My work began in information retrieval, continued through a PhD in computer science at Yale, and moved into fetal MRI at MGH and brain imaging and histology at Penn. Across these projects, I work on extracting useful information from data that can be limited, noisy, or difficult to compare.
questions behind the work
Finding information across languages
My early research studied how to translate search queries when a dictionary offers several possible meanings. I developed an expectation-maximization approach that uses retrieved documents to estimate which translations are useful for a query.
Comparing brain networks across atlases
Brain datasets often use different atlases to define regions, making their connectivity measurements difficult to compare. During my PhD, I developed optimal-transport methods for this problem, including CAROT, which estimates connectomes in a target atlas without requiring the original raw scans.
Locating and segmenting the fetal brain
At MGH, I worked on fetal brain segmentation and extraction from MRI. The method combines models trained with synthetic images and a staged search that first locates the brain, then refines its mask. It addresses limited labeled data and variation in fetal size and position.
Extracting the brain in animal imaging
At Penn, I developed PIGSKIN with collaborators to separate brain tissue from surrounding structures in pig MRI. The method uses synthetic training images derived from a small set of annotated scans, adapting brain-extraction tools to porcine anatomy.
Measuring sparse axonal pathology
With PIGMENT, I am developing a deep-learning framework to segment APP-positive axonal pathology in porcine tissue sections. The work focuses on small, fragmented targets and on measuring pathology beyond pixel overlap alone.
Connecting tissue sections with MRI
I also contribute to workflows for aligning histology with MRI. The goal is to place microscopic tissue measurements in anatomical context and study their relationship to MRI measures of brain injury, including diffusion imaging.
building Scoutics
Alongside my research, I’m building Scoutics: a simulation-driven approach to sports intelligence. The project brings sports data, predictive modeling, and match simulation into one workspace.
The direction is to move beyond a single prediction: let people explore how a game might unfold, change assumptions about lineups or tactics, and examine the range of possible outcomes. Soccer is the starting point for this work, with fantasy football as another application of the underlying models.