PhD Fellow in Bayesian Deep Learning
Former Research Associate at Columbia University
I'm a open minded, project oriented PhD Fellow in Bayesian Deep Learning (previously MSc in Electronic Engineering and MSc in Computer Engineering), always looking for new and exiting challenges. I love working out of my comfort zone, dealing with new environments and heterogeneous projects. My current research interests include Bayesian Machine Learning and Deep Learning and their potential to be implemented on high-performance systems and on custom hardware (notably on FPGAs).
Columbia University in the city of New York, New York (US)
The project will involve the high level design of an accelerator for ConvNets targeting an application-specific domain. The project is a joint collaboration inside Columbia University between the group of System Level Design at Computer Science and the NEVIS Group at the Department of Physics.
Jul. 2017 - PresentEURECOM, Sophia Antipolis (FR)
In the context of the semester project at EURECOM, I spent five months part-time working on a research project on Deep Learning for Unsupervised Learning in the Department of Data Science under the supervision of Prof. Maurizio Filippone. During this period of time, I improved my expertise in Statistical Inference and, in particular, on Gaussian Processes and on their extension to deep architectures. Since these Deep Gaussian Processes has been limited to the case of supervised problems, during the project I extended the existing framework also for unsupervised learning, notably dimensionality reduction and clustering.
Feb. 2017 - Jun. 2017Università degli Studi di Genova, Genova (IT)
As part of the Final Thesis for the Bachelor Degree in Electronic and Information Technologies Engineering, I spent part of the last semester in the Department of Electronic Engineering (DITEN @UNIGE) with two other students in developing and evaluating methods for Feature Extraction, notably Histograms of Oriented Gradients, under the supervision of Prof. Gabriele Moser. This method has beed adapted to work with remote sensing multispectral VHR images, in the context of urban area analysis, and applied to a real case scenario.
Apr. 2015 - Sept. 2015Master of Science in Engineering - Smart Objects
This program has been based on an interdisciplinary track, merging contributions from Computer Science and Statistics, and addressing numerous applied problems. Through the program, I've got the opportunity to learn basic theoretical frameworks and apply Statistics and Machine Learning methods to many problems of interest, as well as develop the Computer Science skills required to understand, operate and extend Intelligent Systems.
Relevant courses: Machine Learning and Intelligent Systems, Advanced Statistical Inference, Advanced Data Science Topics, Operating Systems, Algorithmic Machine Learning
2016-2018Masters of Science in Electronic Engineering - Embedded Systems
This curriculum has given to me the expertise to design and optimize embedded systems, by integrating system and application aspects. Key topics are the organization of programmable hardware architectures for embedded systems, their software and hardware interfaces, and the implementation techniques in hardware platforms, both integrated and multichip.
Relevant courses: Modeling and Optimization of Embedded Systems, Synthesis and Optimization of Digital Systems, Microelectronic Systems, Computer Architectures, Electronics for Embedded Systems
2015-2017Bachelor Degree in Electronic and Information Technology Engineering
Summa cum Laude (110/110 with honors)
The BSc Degree in Electronics and Information Technologies Engineering provided a solid background knowledge in different fields of Electronic Engineering, Information Technologies and Computer Science.
Relevant courses: Advanced Mathematics, Physics, Digital Systems and Analog Electronics, Processing and Transmission of Signals and Images, Systems Theory, Signals and Systems for Telecommunications
2012-2015This project is mainly focused on design, simulation and synthesis of a 32-bit DLX microprocessor with Branch Prediction Unit and Forwarding Logic. The microprocessor has been design on a low level RTL using VHDL, simulated and tested, synthesized and physically mapped using Encounter (by Cadence). The device has been proved to successfully work with the MIPS32 Instruction Set Architecture (Integer Operations only).
The thesis explores the application of a feature extraction technique, called ”histogram of oriented gradients” (HOG), applied to multispectral VHR images. Three relevant cases, regarding the Amiens (France) urban area, have been studied and the results suggest the potential of HOG as a feature extraction tool for urban area mapping and the opportunity to use it together with other feature analysis techniques.