Physics-Informed Neural Networks for Propulsion Simulation May–Sep 2024

I designed and trained a Physics-Informed Neural Network (PINN) using PyTorch, embedding the Navier-Stokes equations directly into the model’s loss function to simulate turbulent flow and predict viscosity gradients.

• Focused on fluid behavior in low-thrust propulsion environments, including MHD flows.
• Tuned hyperparameters (e.g., batch size, learning rate) to improve stability and accuracy.
• Explored the use of PINNs for replacing partial simulation stages in CFD pipelines.

Data-Driven Thermal Protection System (TPS) Optimization Nov 2024–Present

This ongoing project integrates NASA TPS material data with the Materials Project API to explore ablation-resistant materials for re-entry vehicles.

• Built a dataset linking physical properties (thermal conductivity, heat capacity) to ablation performance.
• Used graph neural networks (GNNs) in PyTorch to find material-property patterns.
• Focused on applications for lunar and Martian descent vehicles, where thermal extremes are critical.

This project merges materials science, machine learning, and space vehicle engineering—key domains in my vision for future interplanetary missions.

🛰Autonomous Anomaly Detection for Sensor Networks Jan–Apr 2024

I developed an autonomous system for anomaly detection in spacecraft sensor arrays, targeting real-time failure prediction.

• Built a random forest model in PyTorch, trained on synthetic time-series sensor data.
• Tuned the model to detect performance drops or outliers, simulating spacecraft health-monitoring systems.
• Demonstrated potential reductions in simulated downtime and increased mission reliability.