CV

Project: "Deep Learning-Based Architectures for Semantics Discovery of Entities and Events"

Funded by: HORIZON-MSCA-2021-DN-01-01, project n. 101072488, TRAIL

Supervisor: Prof. Alessandro Sperduti

Modules: Fundamentals of Robotics, Robotics Control, AI & ML for Robotics, Programming Methods, Human-Robot Interaction, Machine Vision, Autonomy in Robotic Systems, Psychology, Ethics, and Standards

Group Project: Smart Hangar Robotics Mobility - Localisation of a mobile robot in featureless environment using C++, ROS, Python, SolidWorks, GitLab

Thesis (Grade 75): Optimized obstacle avoidance using a hybrid navigation framework with YOLOv7, C++, ROS, Linux, GitHub

Research Paper (co-author): Mobile Robot Obstacle Detection and Avoidance with NAV-YOLO. The paper received the Best Paper Presentation award at the 11th ICCMA 2023 conference

Modules: Robotics, Sensing & Perception, Signal Processing, Control Theory, Mechanical Design, Heat Transfer, Electrical Network, Materials

Robotics Project: Music-playing robot with velostat tailor-made keyboard (ROS, C++, Ableton, Arduino, GitHub, Linux)

Multidisciplinary Project: Electrical lunch box (C++, Arduino)

Project: Multimodal Home Assistant for Grasping and Manipulation. Designing a framework for the AI2-THOR default robot to perform various tasks requiring navigation and interaction with objects via grasping in a simulated environment (AI2-THOR) based on vocal commands. The project focuses on combining audio and visual inputs for perception, reasoning, and manipulation, with an emphasis on designing a robust and transparent system to prevent "silent failures" in the voice component during complex grasping tasks. | Supervisor: Prof. Dr. Stefan Wermter | Collaborator: Julia Gachot

Enhancing CoPAL with Explainable Multimodal Summaries: Extending the CoPAL (Corrective Planning of Robot Actions with Large Language Models) framework to generate natural language explanations for robotic decision-making. The project integrates Grounded-SAM-2 for visual processing, hierarchical summarization algorithms, and Qwen 3 LLM to transform complex robotic decisions into clear, human-readable explanations. This research addresses the critical challenge of robot transparency, enabling better human-robot collaboration by making AI decision-making processes interpretable and trustworthy. CoPAL Original Paper | Supervisors: Frank Joublin , Antonello Ceravola

• Analysed the productivity of different assembly lines for robotics applications in a factory in Riga (Latvia), targeting labour productivity and increasing the capacity of the production cell
• Led the study of collaborative robot implementation in production lines. The leading team was delighted and decided to continue the study
• Consulted the team by introducing other industrialisation options such as the creation of a neighbouring cell or the implementation of industrial robots instead of cobots, resulting in a human-robot collaboration
• Proposed a new production cell configuration with possible incorporation by the manager. Studying the different scenarios taking into account the technical, logistics, ergonomics safety and cost drivers

• Took part in the robot building from scratch using materials at disposal such as wood, syringes or plastics. The first robot was remotely controlled to lift the second one
• Collaborated on three Arduino programs: one to manage the servo-motors, another to lift objects and the last one to grab the targets