I'm Pablo Magariños, an Aerospace Engineer and AI researcher. I recently started my PhD in Aerospace Technology at the Universidade de Vigo, focused on embodied AI: how an agent can learn to act competently in the physical world from its own experience of it. I do research with ATRG (Aerospace Technology Research Group) and have just started teaching at the School of Aeronautical and Space Engineering.

My work sits in reinforcement learning and meta-learning under partial observability, applied to multi-agent systems and embodied AI. I study the role that world models and the belief state play in an agent's ability to adapt to novel situations and, eventually, learn from them.

The goal is intelligence that genuinely understands its environment, rather than merely identifying statistical correlations. Drones and aerial robotics are my anchor domain, but embodied AI is cross-cutting: progress here spreads everywhere else.

New environments. New teammates. New knowledge.

• Weakly Supervised Segmentation of Macroalgae Through Gradient Analysis in CNNs and SAM, Applied Sciences (MDPI), 2026 · Paper
• Real-Time Aerodynamic Airfoil Optimisation Using Deep RL with PPO, Aerospace (MDPI), 2025 · Paper · Code
• MSc thesis: Multi-Agent RL for Drone-Based Search and Rescue, architectures, stability and genuine collaboration vs. memorisation

🌐 Website · 🎓 Google Scholar · 🆔 ORCID · 💼 LinkedIn · 𝕏 @pablodotmd · ✉️ Email