Brief: Extend an ML-based reduced-order model (ROM) that predicts surface Cp and wall shear stress on a drone geometry by adding weather/atmospheric inputs and an uncertainty‑quantification (UQ) layer with calibrated confidence estimates. Work alongside the PFE student who built the ROM.

Goals and responsibilities:

• Include weather variables (wind magnitude/direction, turbulence intensity, atmospheric profile) in the ROM input space; curate training data slices from available CFD runs.
• Implement and compare UQ methods (e.g., deep ensembles, MC dropout, calibrated regression); calibrate and benchmark.
• Evaluate generalization across weather conditions and define the reliable operating envelope; deliver a technical note and clean, reproducible code integrated with the existing repo.

Required skills:

• Solid Python and PyTorch; experience with regression on spatial/field data.
• Good ML engineering practices (logging, checkpoints, reproducibility). Nice‑to‑have: UQ, basic CFD/aero, HDF5/VTK.

Planned training:

• Direct supervision by a PFE student; onboarding to codebase and CFD dataset conventions; short readings on UQ for regression.

Other details:

• Targeting students entering their final year and interested in pursuing a PFE with ReDX.
• Recommended period: 3 months.
• Compensation: Monthly stipend.