Projects

Projects

Problem: Building HVAC systems waste energy due to lack of real-time occupancy data.

Solution: Developed a deep learning model using thermal imagery to detect occupancy and optimize HVAC usage.

Tech Stack: Python, OpenCV, CNNs, thermal sensors

Impact: Improved energy efficiency; Deployed in the HVAC Lab at the Graham Hall of NCAT; accepted for oral presentation at IEEE AIPR 2020 and conference paper published.

Links: GitHub

Problem: Schools and public spaces needed fast, contactless fever detection during COVID-19.

Solution: Worked with a team to build a low-cost infrared thermal imaging system for mass fever screening.

Tech Stack: Raspberry Pi, Python, OpenCV, thermal cameras

Impact: Deployed in the Artificial Intelligence and Visualization Lab at NCAT; featured in Fox 5 Atlanta, Fox 8 Greensboro, and PBS North Carolina.

Links:
Fox 5 Atlanta
Fox 8 Greensboro
PBS North Carolina

Problem: Real-world CPS attack datasets are limited and lack variability.

Solution: Created synthetic datasets using GANs and explored LLM-based pattern generation for anomaly detection.

Tech Stack: PyTorch, GANs, scikit-learn, LLM fine-tuning

Impact: Enhanced detection rates and improved model generalization across domains.

Links: GitHub

Problem: Extracting structured metadata from scientific articles is labor-intensive and error-prone when done manually.

Solution: Built a natural language processing (NLP) pipeline and developed a graphical user interface (GUI) application to extract citations, affiliations, and keywords from academic PDFs with high accuracy.

Tech Stack: Python, Fuzzy wuzzy, regex, Tkinter (GUI)

Impact: Deployed for researchers under the NSF Engineering Research Center (ERC) for Revolutionizing Metallic Biomaterials; streamlined metadata extraction for NSF and ARL grant reporting.

Links: GitHub

Problem: CPS face evolving attacks that evade single-model detection.

Solution: Designed an ensemble framework combining ML models and feature selection for robust detection.

Tech Stack: XGBoost, Random Forest, PCA, Python

Impact: Achieved high detection accuracy in multi-domain CPS attack scenarios.

Links: GitHub

Achievements & Impact