I have solid experience in developing, leading and managing postgraduate and undergraduate computer science HE programmes, particularly in the areas of AI and Data Science. I am actively researching in the areas of Deep Reinforcement Learning, Neural Network and Machine Learning with hands-on applications and publications in image processing, autonomous vehicles and robotics. Autonomous learning from data is second nature to me; particularly I enjoy developing reinforcement learning algorithms, several of which have both a highly practical and theoretical potential. I have developed a tool that was adopted throughout a university for quantifying and accurately modeling the data form modules and course outcomes, these algorithms and research have a high potential for commercialisation. I have led a small research group in AI and Robotics along with a few programmes in Data Science, Intelligent Systems and Information Technology for both the undergraduate and graduate levels. I have taught numerous modules, the majority of which in AI and Machine Learning.
My main research area is Deep Reinforcement Learning applied to Intelligent Agents. I have developed several novel RL and DRL algorithms that seamlessly integrate experience replay within the learning process of the agent. I have applied these on games, stochastic processes and on predicting control signals for a prosthetic limp. I have developed several robot navigation models that employed deep reinforcement learning to visually allow a robot to learn to navigate in an unknown environment without prior experience.
Deep Reinforcement Learning on Games
This project aims at building a prototype of a reinforcement learning controller that uses deeply learned features of grabbed frames of a game as input. The agent will be trained to come up with the best control policy to win a game, whether playing against a human or another agent or even self-play! You will be using PyTorch along with an Attari simulator in Python to train your model to try and beat the performance of other models.
Reinforcement Learning for Planning
This project aims at building a reinforcement learning technique that is suitable for planning. The objective is to come up with a set of sub-goals to navigate in a simulated environment from one location to another without storing an explicit map of the environment. The planning behaviour would be explicit through interaction with the environment.
Reinforcement Learning on Simulated Robot Homing
This project aims at developing and training an agent controller that is equipped with reinforcement learning techniques to learn an optimal policy that allows the agent to go back to its home regardless of where it is. Particularly, when the goal is hidden. The idea is to mimic the ability of animals/ insects of reaching its home after foraging for food. This will be done using visual clues and the agent will be left to automatically learn to reach the home location with no map. The environment can be a simulated indoors or outdoors environment and you are free to choose its settings ( factory, flat etc).
Visual Scene Identification using Deep Learning
This project aims at creating and training a deep learning model that is effective in recognising a scene even when the angle is different than the one the model was initially exposed to. You will use a dataset that is created via a simulation or real scene images. Each image would have taken from a set of different angles and a fixed distance (or from an omnidirectional camera). The model should be trained to achieve high accuracy of identifying a scene from a reasonable distance and from different angles.
Emotion Recognition using Deep Learning on a GPU
This project aims at creating and training a deep learning model that is effective in recognising human emotion from different angles. You will use a readymade dataset of pictures taken for individuals expressing seven emotions - happy, sad, angry, neutral etc. Then the model should classify an image as expressing one of the aforementioned emotions. You will build the architecture using PyTorch or TensorFlow or any other deep learning framework and you will aim at deploying and training the model on a multi GPU server
Gun Detection from a CCTV Feed using Deep Reinforcement Learning
This project aims at creating and training a deep learning model that is effective in recognising a gun in the hand of a person in a CCTV setting. You will be using the YOLO framework with its ideas that stemmed from CNN (Convolutional Neural Networks), R-CNN, Fast R-CNN and Faster R-CNN. You would need to identify a suitable dataset other than the COCO dataset.
- PhD in Reinfrocement Learning
- Master in Artificial Intelligence
- Higher Diploma in Informatics
- BSc. in Mathematics-Informatics
- IEEE Intelligent Systems Society
- IEEE Robotics and Automation Society
- Artificial Neural Networks Society
Currently, I am developing several modules for the new MSc. in Data Science ODL programme. I lead modules such as Data Science, Reinforcement Learning on Robotics, Machine Learning, Neural Networks and Big Data Analysis. I have served as a Programme Director for MSc. Data Science and Computational Intelligence, where I enjoy setting up suitable strategies for this programme to be closely related to both the industry and research (my favourite balance), nurturing it into popularity and growth. I have obtained managerial skills by undertaking different leading roles and responsibilities, ex. appointed as acting dean of my college and coordinating BSc. in Business Information Technology program for three years. My experience of educating professional students includes proficiency in teaching several AI and computer science topics in multicultural settings. Extensively proposed and championed several curriculums in Data Science and AI at the Master and Undergraduate levels.