Farooq Ahmad Wani
dottorando
telefono:
+393482293812
Biography
I was an Artificial/Machine Learning Engineer with four and a half years of experience, where I specialized in designing, implementing, and optimizing various machine learning models at General Electric (GE). My work involved deep engagement with data science and its related fields, including Deep Learning, Machine Learning, Artificial Intelligence, and Programming, all of which I found immensely stimulating and motivating.
Currently, I am pursuing a PhD in Data Science at Department of Informatica, Automatica, and Gestionale (DIAG) , La Sapienza, Rome. Now in the second year of my PhD, I am conducting research under the supervision of Professor Fabrizio Silvestri. My research primarily focuses on Graph Neural Networks (GNNs), particularly on enhancing their efficiency when dealing with graph data that suffer from limited and noisy labeling.
Beyond GNNs, my research also explores strategies for training models in the presence of noisy data, which includes developing robust models, techniques, and loss functions, as well as understanding how models learn in such environments. Additionally, I am interested in how overparameterized networks can be leveraged to perform multiple tasks simultaneously, which is another key area of my research.
Research interest
I have a broad range of interests within data science, particularly in machine learning, artificial intelligence, and deep learning. My primary objective is to develop robust methodologies that can effectively handle noisy data during training, ensuring that models remain resilient in such environments. I am focused on designing methods, techniques, models, and loss functions that can withstand the challenges posed by noisy data. While this interest spans all types of training, I am currently concentrating on applying these strategies to Graph Neural Networks (GNNs) in noisy environments.
My work includes in-depth exploration of how training processes unfold in noisy settings across various deep learning models. I have successfully developed a robust loss function tailored for generalized deep neural networks, and I am now extending this work to GNNs. Specifically, I have extensively studied the training of GNNs for graph classification tasks in the presence of noisy labels, aiming to enhance their performance under such conditions.
In addition, I am deeply intrigued by explainable AI and the interpretability of models. I am passionate about developing models that are not only universally applicable across different tasks but can also adapt dynamically to variations in input signals.
Furthermore, I am interested in researching overparameterized networks, particularly in leveraging their additional parameters to perform auxiliary tasks beyond the primary objective. This exploration of how to repurpose surplus network capacity for multitasking is another key area of my research.
Awards
- Winner of TCS hackathon 2015
- General Electric hub award winner 2018
- Highest placed student of my college 2016
- Above and Beyond award winner (General Electric) 2020
- Qualified AIEEE 2012