Curriculum Vitae

Postdoctoral Researcher

National Institute for Research in Digital Science (Inria)

Developing 3D-VLMs for embodied perception and reasoning.

Machine Learning Engineer

Logitech

Developed multimodal deep learning models for emotion recognition in streamers; improved performance via multimodality fusion.

• +6% over SOTA for emotion recognition
• +14% performance improvement through multimodality fusion

Robotics Engineer

OCP Group

Designed a high-fidelity autonomous mining truck simulation and implemented planning and control algorithms.

• Hybrid A* trajectory planning
• NMPC path-tracking

Deep Learning Researcher

VITA-Lab, EPFL

Research on pedestrian future position and intention prediction for autonomous vehicles; published and outperformed prior work.

• Published at hEART 2020
• +4% over state-of-the-art

PhD

National Center for Scientific Research (LAAS-CNRS)

Computer Science (PhD)

MSc

Swiss Federal Institute of Technology (EPFL)

Robotics (MSc)

BSc

Swiss Federal Institute of Technology (EPFL)

Micro-Engineering (BSc)

3D-VLMs for Embodied Perception

Research and development of 3D vision-language models for embodied perception and reasoning.

AI-Based Robot Task and Motion Planning

Graph Neural Network-based geometric reasoning methods for robot task and motion planning (LAAS-CNRS).

Multimodal Emotion Recognition

Multimodal deep learning models combining video, audio and physiological signals for emotion recognition (Logitech).

Autonomous Mining Truck

High-fidelity ROS simulation and planning/control algorithms for autonomous mining trucks (OCP Group).

Graph Neural Network-Based Geometric Reasoning for Robot Task and Motion Planning

ICLR

GNN-based geometric reasoning methods applied to task and motion planning for robots.

Recurrent-Neural Networks for Uncertainty Prediction in Robust Motion Planning

ECAI

Using RNNs to predict uncertainty for robust motion planning.

Generalizing CNN-Based Feasibility Prediction to Multi-Robot Manipulation of Complex Objects

IROS

Extending CNN-based feasibility prediction approaches to multi-robot manipulation scenarios.

Multi-task Convolutional Neural Networks for Joint Action and Grasp Feasibility Prediction

IROS

Multi-task CNNs for predicting action and grasp feasibility in robotic manipulation.

1st Prize in the Virtual Autonomous UAV Inspection Challenge

2022 IEEE RAS Summer School on Multi-robot Systems

Planning collision-free trajectories of two UAVs (Unmanned Aerial Vehicles) in a 3D environment with obstacles, so that cameras onboard the UAVs inspect a set of N unique inspection points.

2nd Prize in the Real-World Autonomous UAV Inspection Challenge

2022 IEEE RAS Summer School on Multi-robot Systems

Planning collision-free trajectories of two UAVs (Unmanned Aerial Vehicles) in a 3D environment with obstacles, so that cameras onboard the UAVs inspect a set of N unique inspection points.

Winner of the iCAN contest swiss selection

iCAN Contest

Design of a portable mobile robot for real-estate mapping and virtual visit.