Research Interests
Welcome to my Website
Ross Brancati, PhD
I am currently a research engineer in the Warfighter Lab at Galvion
More about the Warfighter Lab:
About Me
Recent Highlights
Reinforcement Learning for Stable Gait and Navigation with Quadruped Model
I used Unity to build a quadruped model and trained it with reinforcement learning to walk with fast and stable gait. I also trained it to perform a target navigation task. It uses a central pattern generator and modulates joint angles with a recurrent neural network to overcome internal and external perturbations.
Procedural Animation for a Quadruped Unmanned Ground Vehicle
As a part of the Warfighter Lab's simulation efforts, I developed a procedural animation for a quadruped unmanned ground vehicle. All legs motions are purely animation and do not use physics. This was developed in Unity.
Sliding Puzzle Game
I built a simple 2D sliding puzzle game in Unity and deployed it to a web app. The goal is to slide tiles to create a path between the black dots before the timer runs out. It's free to play at slidesgame.com. Note that it is optimized for play on a mobile device.
Going the extra mile: deployng your ML model
Model deployment isn't in your day-to-day job description? I wrote a guide to a bare bones approach to getting your first model deployed (for free) with FastAPI, Docker, and Render.
Dissertation Defense
I successfully defended my dissertation in January 2025 and officially walked in May. My project explored identifying and classifying patient specific injury mechanisms and factors related to development of patellofemoral pain using unsupervised and supervised machine learning models.
Leave One Subject Out Cross Validation for Machine Learning Models
Cross validating machine learning models by leaving out all observations from a single subject. A great approach for small sample sizes and multiple observations per sample. This post explains the approach and how to implement it from scratch in python.
Gait Speed Classification using Inertial Measurement Units and Deep Learning
Wearable technology such as inertial measurement units are useful for measuring movement. In this project, I used an open source dataset with signals derived from inertial measurement units to classify walking gait speed with a deep learning model.
Principal Component Analysis on Time Series Data
An application of machine learning methodologies including dimensionality reduction and classification models to analyze the complex behavior of musculoskeletal injuries with a "systems theory" approach.
December 2025