Bridging Traditional Engineering with Advanced Analytics

Built a real-time structural health monitoring system applicable to any bridge with existing camera infrastructure. The hybrid approach combines a modified Lighthill-Whitham-Richards (LWR) traffic flow model with Random Forest prediction (R² = 0.88). The system captures live video streaming, processes it using YOLOv8 computer vision for vehicle detection, utilizes Open-Meteo weather API for real-time weather integration, and runs Monte Carlo simulations to quantify uncertainty in fatigue forecasting. Demonstrated through an interactive Streamlit dashboard with maintenance scheduling recommendations.

Developed a climate-aware simulation framework for distributed AI workload routing across 21 global data center locations. The model combines Open-Meteo weather data, a temperature-humidity-PUE energy formulation, a wind-diluted temperature-rise proxy (ΔT-AR), and five routing strategies: random, energy-only, heat-only, ΔT-aware, and multi-objective. A Bayesian-optimized ANN accelerates repeated energy estimation, while 100-trial Monte Carlo perturbations test robustness under weather and demand uncertainty.

Built a disaster management workshop prototype for urban flood risk assessment using Orlando, Florida and Hurricane Ian (2022) as the demonstration scenario. The workflow combines USGS elevation data, terrain analysis, SCS Curve Number runoff modeling, Open-Meteo historical weather, and Random Forest flood classification. Interactive Folium maps, hotspot clustering, road-network routing, shelter screening, and a time-stepped storm replay show how GeoAI and digital-twin concepts can support disaster response planning.

Investigated the correlation between maintainer fatigue and Army Aviation helicopter readiness using 2,865 maintenance records across 37 aircraft. Initial linear regression showed weak correlation, so we transitioned to Random Forest which captured non-linear relationships. Generated synthetic fatigue survey data for 341 personnel using CTGAN from Synthetic Data Vault. Ran 500 Monte Carlo simulations with Gaussian noise to validate findings and quantify uncertainty in workforce allocation strategies.

Designed a Finite State Machine (FSM) Mathematical Model framework for dynamic military unit management integrating real terrain data. Processed USGS Digital Elevation Model data in QGIS to calculate slope percentages and classify terrain into four categories (Flat, Valley, Hilly, Mountainous). Units automatically transition between Idle and Action states based on computed metrics like Force Output, Velocity, and Power. Monte Carlo sensitivity analysis in @Risk identified slope percentage as the primary risk factor affecting unit performance.

Designed and fabricated a functional electric bike with hybrid charging capability (grid power and solar panels). The prototype features a DC motor with gear system, reverse functionality, and regenerative charging concept. Built using recycled materials to demonstrate sustainable engineering principles. This hands-on project integrated mechanical design (SolidWorks), electrical systems integration, and renewable energy fundamentals.