Emanuel Samir Muñoz Panduro

I'm developing safe drone navigation at Carnegie Mellon University's Robotics Institute, using Gaussian Processes, Model Predictive Control, Safe Optimal Control, and Neural Operators to address smoke-impaired visual sensing. My multi-robot, risk-aware path-planning research reduced dense-smoke exposure by up to ~85% versus baselines and was accepted to IROS 2026.

At NoteworthyAI, I built real-time edge inspection systems on Jetson Orin with C++ and ROS2 for utility-pole monitoring, stereo visual tracking for asset localization, and AWS/PyTorch pipelines for production ViT detection and defect classification. I also prototyped OOD detection for Vision Transformers, achieving up to ~70% recall, and fine-tuned long-tailed models to ~90% F1-score for object grid-defect classification.

I've also developed safety-critical autonomous-vehicle planning validated with zero safety violations in CARLA, surgical 3D reconstruction and torque-control systems, and reinforcement-learning and multi-agent coordination projects.