I am a Computer Science and Engineering Master's student at Politecnico di Milano, specializing in Agentic AI applied to Robotics.
My work focuses on bridging spatial-semantic reasoning, real-world scene understanding and human-robot interaction in robotic and embodied systems.
Currently, I am focusing on Large Language Models for Language-Driven Autonomous Robotics for my Master's thesis, with previous work on IoT architectures and SoC firmware development from my Bachelor's Thesis and Formula SAE experience.
- 🎓 Current: Master's in Computer Science & Engineering (Artificial Intelligence) @ Politecnico di Milano
- 🔬 Research: Large Language Models and Semantic-Spatial Memory in Robotic Agents, VLMs for Scene-understanding, Human-Robot interaction
- 🏗 Background: Former Electronics Division Member at UniBS Motorsport & Post-Bachelor Research Intern at the eLUX Laboratory (University of Brescia)
- 🎲 Interests: Embodied AI in Robotics, Natural Language Processing (LLMs/VLMs), Reinforcement Learning
- 🎯 Life Goal: Recreate "GLaDOS" from Portal 2 and summon it in the real world
- Designed a VLM-driven agent architecture using Florence-2 and GPT-4o
- Combined a semantic graph (NetworkX) with a vector database (Milvus) for efficient Retrieval-Augmented Generation (RAG)
- Deployed on an AgileX robot for real-time interaction and autonomous navigation via ROS 2
- Developed ESP32 firmware for CAN-Bus / OBD-II data logging
- Built an MQTT gateway for real-time telemetry transmission
- Created a Python OBD-II emulator using Assetto Corsa to simulate vehicle sensor data
- Built a computer vision pipeline using YOLOv8 and OpenPose
- Performed pose estimation and ball trajectory tracking on tennis match footage
- Applied LSTMs, Transformers, and embeddings to chess datasets
- Compared classical models with Zero-shot and Few-shot LLM approaches for move prediction
- Email: paolo13.riva(at)mail.polimi.it
- LinkedIn: linkedin.com/in/rivapaolo/