PINNs & SciML Laboratory on AWS
Physics-Informed Neural Networks for PDEs (1D Burgers, 2D thermal diffusion, Navier-Stokes) with FastAPI microservices and AWS EC2 deployment.
Project Overview
This ecosystem of projects applies Scientific Machine Learning (SciML) to classic and advanced problems in mathematical physics. Fundamental conservation laws (mass, momentum, energy) are directly embedded into the network’s loss function via automatic differentiation.
Modeled Equations & Cloud Architecture
- 1D Burgers Equation & 2D Thermal Diffusion: High-precision numerical modeling of shockwave fronts and transient heat conduction profiles.
- Navier-Stokes Fluid Flows: Non-linear incompressible fluid simulation with inverse drag coefficient estimation.
- AWS Cloud Infrastructure: Microservices built with FastAPI, containerized via Docker and orchestrated on AWS EC2 compute instances, integrating persistent model checkpoints in AWS S3 with strict IAM security policies.