Alyssa Perry

Ai Ml Engineering Intern @The Aerospace Corporation

El Segundo, CA, US

SHAP for Reinforcement Learning Interpretability:• Developed a reusable SHAP interpretability pipeline for RLlib PPO agents, resolving API and model integration issues.• Engineered custom PyTorch wrappers and logit isolation to enable SHAP DeepExplainer compatibility across RL agents.• Collaborated across teams to debug, optimize, and deploy the SHAP pipeline on a 13-action policy with 8 features.• Validated the workflow on a CartPole PPO agent before scaling to the RPO simulation framework.Synthetic Imagery Quality Evaluation:• Designed a scalable Python pipeline to evaluate sim-to-real gaps in satellite imagery using Fréchet Inception Distance.• Built PyTorch-based modules using Pillow and NumPy for large-scale image sampling, resizing, and transformation.• Delivered results to stakeholders, shaping evaluation standards for synthetic training data in missile detection models.

Chapman University

Bachelor of Applied Science - BASc, Computer Science