Industrial CV
Transfer learning classifiers (ResNet50, Xception) reached 99%+ accuracy on wood chip identification.
AI/ML Researcher ยท Starkville, MS
I love AI research and building infra that turns vision models into real-world systems.
B.S. Computer Science student at Mississippi State University. Focused on computer vision, semantic segmentation, and efficient deployment.
Now
Mississippi State University
B.S. Computer Science, Data Science minor
Aug 2024 - May 2028
GPA
4.0 / 4.0
Focus
CV, ML, Systems
Highlights
- 99%+ accuracy in wood chip classification
- U-Net segmentation with elastic augmentation
- YOLO Nano for CubeSat fire monitoring
Focus areas
Computer Vision
Semantic Segmentation
Efficient Inference
Reinforcement Learning
About
Research-driven, deployment-minded.
I build end-to-end AI systems that move from experiments to dependable tools. My recent work spans industrial computer vision for forest products and reinforcement learning for wireless communication.
I care about model efficiency, reliable data pipelines, and evaluation that holds up outside the lab.
Coursework
Data Structures and Algorithms, Software Development, Database Management Systems, Computer Networks, Linear Algebra, Machine Learning, AI, Data Wrangling, Probability, Systems Programming.
What I care about
Reliable training pipelines, efficient inference, and real deployments where accuracy, latency, and robustness meet.
Highlights
Measurable wins and systems wins.
Segmentation Pipeline
U-Net with ResNet34 backbone, elastic transforms, and Dice loss for robust tree ring detection.
RL + NLP
DistilBERT embeddings inside a hybrid actor-critic model improved wireless optimization performance.
Experience
Applied research and engineering.
Researcher Intern - AI2F
MSU Department of Industrial and Systems Engineering
- Designed and deployed transfer learning classifiers (ResNet50, Xception) to identify Southern Yellow Pine and Red Oak wood chips with 99%+ accuracy.
- Built a semantic segmentation pipeline in PyTorch using U-Net with a ResNet34 backbone, elastic transformations, and Dice loss optimization.
- Engineered test-time augmentation and weighted sliding-window inference, applying Zhang-Suen thinning to refine output masks.
Undergraduate Research Assistant
MSU Wireless Communication Research
- Adapted deep reinforcement learning algorithms to optimize network parameters and stabilize convergence in stochastic environments.
- Fine-tuned a DistilBERT large language model within a hybrid actor-critic network using semantic embeddings.
- Developed a GPU-optimized PyTorch training pipeline with custom experience replay buffers for large-scale matrix operations.
Projects
Systems built for constraints.
CubeSat Forest Fire Detection
Lightweight YOLO Nano (2.3M parameters) for resource-constrained CubeSat fire monitoring using PyTorch and AMD ROCm.
Neural Network From Scratch
A custom deep learning library in Python with a tensor engine and foundational linear algebra operations.
Skills
Tools I build with.
Languages
Python
C/C++
SQL
MPI
Libraries
PyTorch
TensorFlow
Keras
TensorRT
NumPy
Pandas
Scikit-Learn
Matplotlib
Tools
Git
GitHub
Linux
MLOps
Firebase DB
Selenium
BeautifulSoup