
SeatSwap: Physics-Grounded Generative Motion Adaptation
Meta Reality Labs · SIGGRAPH 2026 — Poster, Selected for Technical Paper Track
Applied Research · Physical AI · Agentic AI
I develop AI methods and translate them into scalable systems — across simulation, perception, and production. My work spans computational biomechanics, wearable sensor perception, scientific machine learning, agentic AI, and production character animation. I build and harness AI systems end-to-end: from research prototype to deployed pipeline.

About
I work at the intersection of simulation, perception, and production — developing AI methods and systems for understanding human motion and physical interaction. My background spans a PhD in Computer Science and Biomechanics from Iowa State University, postdoctoral research in scientific machine learning at the University of Notre Dame, and current work in computer vision and physical AI at Meta Reality Labs.
My research experience includes wearable and camera-based sensing, time-series analysis, temporal deep learning, and human–object interaction. I'm also interested in how generative and agentic AI can support robotics and embodied systems.

Now
Computer Vision Engineer II
Physical AI · Meta Reality Labs (Contract)
Production
Postdoctoral Research Associate
Scientific Machine Learning
University of Notre Dame · AI & SciML
Simulation
Doctoral Graduate Researcher — AI
Computer Science & Biomechanics
Iowa State University
Perception
Selected work
Applied AI & Production Systems

Meta Reality Labs · SIGGRAPH 2026 — Poster, Selected for Technical Paper Track
Applied AI · Fine-Tuning
Applied AI · RAG
Applied AI · Agentic Systems
Perception & Wearable Sensing

AAAI 2024 Spring Symposium on Clinical Foundation Models

Computer Vision · Sports Analytics

Sensor Fusion · IMU
ML Foundations & Data Science

Applied ML · Anomaly Detection
Gaussian Anomaly Detection — Network Infrastructure AI

Applied ML · Recommender Systems
Neural Collaborative Filtering — Personalized Recommendation Engine

AI · Game Theory
Adversarial Game AI — MinMax Decision Engine

NLP · Political Data Science
Political NLP — Tweet Sentiment & Electoral Trend Analysis

Computational Biology · Dimensionality Reduction
Multi-Omics ML — Manifold Learning for Gene Perturbation Networks
Generative Models
Three generative model demos running entirely in-browser — zero API calls, zero server, fully static. Each implements a real algorithm from scratch.
01 · Inverse Kinematics
FABRIK Real-Time Solver
Move your mouse. 7 joints solve for it in real time — the same algorithm behind character animation rigs and robot arm control.
02 · Discrete Diffusion
Text Denoising Animation
Characters start as noise and forget what they are. Cosine variance schedule (Improved DDPM), no neural network.
03 · Markov Chain
Research Title Generator
Trained on your papers. Bigram model, 358 states — different title every run.
Publications

SIGGRAPH 2026 — Poster, Selected for Technical Paper Track
SeatSwap: Learning-Assisted Differentiable Kinematics for Human-Object Interaction Adaptation in Production-Scale Character Animation
APL Electronic Devices
Scalable Non-Diffusive Thermal Analysis of Field-Effect Transistors Using Neural Operators
Materials Today Physics
Physics-Informed Neural Networks for the Phonon Boltzmann Transport Equation with Hard-Encoded Angle-Dependent Boundary Conditions

AAAI Spring Symposium on Clinical Foundation Models
Striding into Clarity: Wearable Sensor-Driven Estimation of Knee Adduction Moment, Unveiling the Black Box with Sequence-Based Neural Networks and Explainable AI

Wearable Technologies
Center of Mass Acceleration During Walking: Comparison Between IMU and Camera-Based Motion Capture Methodologies

World Scientific Annual Review of Biomechanics
Assessment of Gait Balance Control Using Inertial Measurement Units — A Narrative Review

Archives of Gerontology and Geriatrics Plus
Estimation of STEADI Performance Using Inertial Measurement Unit

IX World Congress of Biomechanics
Estimation of Knee Adduction Moment During Walking Using Wearable Sensor Data With Sequence-Based Artificial Recurrent Neural Network
Recognition
Honors & Awards
Leadership & Service
Full project archive

View Code Piplines LLM MLLMs sentiment analysis Audio Text Multimodalities google vit base patch16 224 BERT

View Code RAG embeddings Huggingface LLM context aware chat engines memory buffers

View Code Multimodal LangChain OpenAI whisper speech to text AI chat engines AI agent
Developed anomaly detection algorithms to identify abnormal server behavior using throughput and latency metrics. Applied Gaussian models to visualize and detect anomalies in 2D datasets and extended the approach to high-dimensional data, achieving high detection accuracy through cross-validation. Optimized detection thresholds using precision-recall metrics, improving the accuracy of identifying true anomalies.
View Code Python Matlab sklearn Multivariate Gaussian High Dimension
Developed and optimized collaborative filtering algorithms to predict movie ratings using a dataset of 1682 movies and 943 users. Implemented regularization techniques to improve model generalization and minimize overfitting. Enhanced user satisfaction by personalizing movie recommendations based on user preferences, improving engagement.
View Code Python Matlab sklearn Collaborative Filtering Regularized Gradient
A program that read IMU signal to estimate the real-time orientation and reconstruction of the object’s orientation back to 3D space.
View Code MATLAB
This project is an implementation of the classic Checker game with an AI opponent that utilizes the MinMax algorithm for decision making. The MinMax algorithm is a widely-used technique in game theory for determining the best move in a two-player, zero-sum game.
View Code Java javascript Min Max
Developed real-time pose estimation algorithms using OpenPose to track dynamic gymnastics movements. Designed data structures for body points and trajectories with interactive visualization tools, ensuring accuracy and usability. Enhanced algorithms and user interface through testing and feedback from gymnasts and coaches
View Presentation Python OpenPose sklearn OpenCV Caffe
Developed a Java application integrated with SQL to analyze state legislators’ tweets during an election year. Designed and executed SQL queries to extract insights from tweet content, trends, and sentiment. Provided data-driven visualizations to understand the political discourse and its evolution over time.
View Code Java javascript SQL Tweeter JDBC
Python javascript PyTorch Jupyter sklearn Scikit Optimize pandas NumPy
Python Java PyTorch Jupyter sklearn javascript SQL XG Boost pandas NumPy seaborn SciPy
Python R GENIE3 t SNE UMAP PCA
Java
Machine Learning with Membership Privacy using Adversarial Regularization
Comprehensive Privacy Analysis of Deep Learning:Passive and Active White-box Inference Attacksagainst Centralized and Federated Learning
Adversarial Example in Biomech Research
Cloud Computing in Healthcare and Privacy
Classify individual with Overweight by IMUs Signals
Situation Awareness in Falling Prediction and Prevention
AI Survey Over Applications of Patient with Stroke
Advanced Algorithm
Theory of Computation
Advanced Computation (class content is not allowed to upload)
Turing Machine (TM) and Prove Computability, Decidability, Countability, Slice and Parameterized of Languages, Enumeration, Universal TM, Prefix TM, Classes and Subset, Quantum Computing
University Research Excellent 2024 View Certificate
“The purpose of this award is to recognize the top 10% of graduate students for outstanding research accomplishments as documented in their theses and dissertations.”