Development of production-ready deep learning models to be run on cloud and edge devices. Thisincludes data collection, annotation and processing; model definition, training, optimization, profiling,testing and developing the infrastructure for inference using AWS and Gstreamer. Some of the projectsI spearheaded were:- Action recognition and localization.- 2D/3D Player and ball detection and tracking, from monocular images.- Advanced match analytics using AI

Re-implementation and improvement of state-of-the-art methods for Deep Learning in Computer Vision tasks. These include:- Object recognition, detection and segmentation (one and two stage)- Human action recognition and detection- 3D Human Keypoint Estimation- A little bit of GANs within the SuperResolution domain- Everything was implemented with Python, Tensorflow and Pytorch- Supervision of Interns

Supervision of 3 Master Thesis:1. A Revolute Joint Module for the Fable Robotics System by Mathias Grelck Johnsen (Collaboration with ShapeRobotics)2. Obstacle detection and avoidance for indoor mobile robots using Recurrent Neural Networks by Szymon Prezemyslaw Kowalewski3. Transfer learning between Neural Networks by Felix Johann Sebastian Laumann

The project focuses on the seamless integration of AI and CT through :1. Developing computational and control models of detection and pick-and-place tasks for quotidian object manipulation in human environments.2. Design and development of scenarios in which the model can be tested: human-robot interaction, object recognition and force-feedback manipulation.3. Analyzing the impact of this merged model in aspects such as flexibility, robustness and adaptability and comparing the results to each individual approach.Research topics:- ROS (Robot Operating System)- Robot/Object Interaction- Convolutional Neural Networks (training and implementation)- Object segmentation (Semantical and geometrical)- Mobile/Humanoid/Service robotics- Task intelligence through Episodic memory (Deep-ART)- 3D point cloud processing (PCL Library)- Object gasping and manipulation- SLAM

Teaching Assistant on the course: 31383 Robotics .Covered Topics: - Introduction to Robotics - Robot Forward and Inverse Kinematics - Velocity Kinematics and Trajectory Planning - Manipulator Dynamics - Robot Control - Sensor system in Robotics

Supervision of 2 Master Thesis:1. Task constrained motion planning for robot manipulators by Marie Claire Capolei2. Realtime motion planning for industrial robot arm in dynamic environments by Carlos Corchado

Research Assistant for PhD project on Automation of Combine HarvesterTopics:- Sensor fusion- UAV vision- Image processing- Control Theory