Business Intelligence and Incident Analisys
The project is commissioned by the MPS Bank and it aims to apply Business intelligence techniques including data science and Machine Learning to upgrade some of their internal process, control and incident management systems.
The bank system was gradually created during years by different modules and departments. The information systems and the data structure is not properly designed for an homogeneous architecture.
Alert and predict suspicious tickets that may be effects of underlying system novelties by using LSTM architectures and proper incidents catacterization.
SAme approach can be folled for the software release scheduling to anticipate risky releases, or combinations of them, that may cause problems. Manage the internal Process and Controls organization related with KPIs.
Technologies: Tensorflow, Pandas, Numpy, LSTM
Reinforcement Learning and Curiosity
Industry 4.0 as a Service. INASSE Framework
INASSE Industry 4.0 as a Service - Design of an IoT framework composed by HW, SW, Cloud and ML modules to favor Small and Medium Enterprises to approach Industry 4.0.
IN.AS.SE is a project founded by the Toscana Region. It aims to create a framework of resources that allow SME to approach the world of Industry 4.0.
It aims to reduce the impact on SME of the production transformation and the transition of the enterprise to Industry 4.0.
The whole set of resources will be available to all the customers “as a service” according to the most recent trends in terms of product philosophies. Each customer won’t be charged the full price at the moment of contract sign. The cost of all the elements of the framework (Hardware, software, cloud) will be spreaded on the whole operating time, reducing the initial financial step.
All the framework elements will be designed to work in a sinergic and complementary way, empowering the direct presence of the hardware productor SECO and the software service creators.
UDOO Boards - Design, develop, test, disseminate a family of Enabling Technologies (HW & SW) for Digital DIY and industrial fast prototyping.
UDOO was a visionary project by the hardware manufacturer SECO and Siena university that aimed to develop a set of enabling technologies for DIY, education and hobbyist which evolved to an advanced prototyping tools set for startups and companies.
UDOO was the first ever small sized computer that merges in one single board an ARM cortex-A9 iMX.6 CPU and an Arduino Due compatible board embedded with a dedicated ARM SAM3X8E CPU.
UDOO merges different computing worlds together: educational purposes, Do-It-Yourself (DIY) and rapid prototyping. During the years different models and editions were released.
Identify key features, use case scenarios, hardware/software feature design, Linux and Android BSP Development, documentation, tutorials, marketing and dissemination.
Off Grid Box
The INASSE project included an internal test field to verify the validity of the IoT enabling technologies developed. In the specific case, this concerned the transformation of the Off Grid bBox into a smart device. The OGB is a device for generating, storing and managing solar energy and desalination and purification of water in off-grid areas. A selection and interfacing of additional sensors, a specific docker distribution, the creation of a cloud service and some ML modules for the management of the autonomous box in case of lack of connectivity were carried out.
FIND - Taking Machine Learning into a Steel Mill
FIND – Design an IIoT solution to transform an old steel mill plant into a Industry 4.0 ready device by a Retrofitting approach based on Computer Vision and Machine Learning.
The main focus of this project was the Industrial Smart Retrofitting activity.
The transition to Industry 4.0 has not the same impact on every production sector. For instance, the upgrade process in a steel mill is much more complicated than in a web agency. Heavy machinery has a much longer time span then other activities. In this project we aimed to solve some bottlenecks in a steel mill production process: retrofit old machinery to predict failures in mold production phases, use an external machine learning module based on computer vision to control the quality of production to avoid long time blocks due to production issues. All the process was based on Design thinking methodology in order to involve the whole production stuff, owner, management, operators.
The final goal was to integrate all the modules in a user device for the operators, and not only, in order to maximize the benefits for all the involved users.
SEAL - SEco Assembly Line
SEAL – Develop an Assembly line visual Machine Learning system for quality control.
The embedded hardware manufacturer company SECO is required to verify the quality of the pieces at the end of the assembly line. The manufacturer has a fee for every piece delivered with an issue and, for sure, a loss of trust. The project aims to implement an image detection system that identifies novelties and allow the user to be part of the continuous system improvement and to be a part of an IIoT infrastructure following the LEAN philosophy. THe deployment consists in system design, user interactions, model deployment, system integration in the production line. The whole process was developed applying design thinking methodology.
Technologies: TensorFlow, YOLOv3, Python
UX 4 CLEA
CLEA - Seco IoT Platform
IoT platforms have reached a crucial role in the IoT effective usage and diffusion: Setup, device management, data and business intelligence, maintenance, divide updates, customized UI.
I participated in the definition of key features, built the model of the system, organize the entities and design the interfaces for the SECO CLEA platform.The experience I had in IIoT projects and applications deployment allowed me to share my knowledge with the pure design team and identify the needs for the different final users.
UAPPI - UDOO AppInventor
UAPPI, AppInventor - Design and build a AppInventor distro to allow visual prog)ramming in physical computing based on UDOO boards.
The aim of the project is to develop a mod for the famous visual programming platform for Android applications, Appinventor, which allows the use of UDOO rapid prototyping boards. To allow the use of Arduino resources (sensors, actuators) to favor the practice of Physical Computing through Android applications in order to expand the resources available for introducing students to computational thinking.
Prevent ATM Bank Attacks
Design a Machine Learning system for ATM physical attack predictions based on Intel Real-Sense 3D camera.
Attacks on ATMs have become more and more frequent after the availability of cash in bank branches is increasingly reduced. The ATMs are equipped with various defense systems which, in the event of tampering, activate alarms and mark the banknotes making them useless. The attacks are often very invasive and endanger the very structure of the building and cause costly damage. The ATM rooms are monitored in a control room with hundreds of screens impossible for operators to monitor.
The system, developed on order from MPS, aims to identify anomalous behaviors (use of burglary tools, and report pre-alarms to the control room in order to activate dissuasive and preventive measures.
The system is based on depth chambers (in order to respect privacy) and deep learning algorithms and gaussian mixture model, proposing to identify patterns on user behavior by dividing them into normal and anomalous.
PwC - Office IoT for Wellness
UDOO at PwC - Design and Build an Iot solution for a Office Wellness Environment at Heinz College & PwC Risk and Regulatory Services Innovation Center – Pittsburgh PA (USA).
The project explores the possibility of deploying a Internet of Things (IoT) testbed in PwC’s New York City office to explore and evaluate benefits to all stakeholders (building owner and manager, employees, and PwC) and help understanding security and privacy implications, and help inform the adoption of best practices and needed research.
To Increase the extrinsic motivation for employees to follow a healthy lifestyle in order to improve the employers wellbeing, link with the company, and productivity.
To fit these objectives we designed and prototyped a gamified system that promotes good and healthy behaviors. At the same time it can improve the efficiency of buildings and it aims to provide the following benefits for respective stakeholders:
PwC and office management can monitor office dynamics in an anonymous way
Employees save time and avoid stress for banal actions
PwC’s workforce enjoys more health, more happiness, and ... more productivity
The company benefits from a reputation of being innovative and forward-looking, internally and externally.
Technologies: Python, Hardware prototyping,, UDOO Boards
High school teachers and students, teaching institutions, researchers, stakeholders and actors on education and learning methods and systems, experts and industry companies or individual professionals in Ubiquitous, Mobile and IoT (UMI) technologies, career consultants and policy makers are welcomed to the Communities of Practice (CoPs) that are developed within the context of UMI-Sci-Ed project.
FILME Interactive Lighting
This project aims to create a set of lamps for indoor lighting and audio entertainment systems. After a preliminary market analysis of user requirements through Design Thinking methodologies IThe task was the design of a custom touch control surface based on an ultrasound system to transform surrounding surfaces to control areas and a set of interaction prototypes based on LED light effects and custom touch surfaces.
Indole smart power plug
Museums Iteractive Installations
Build a set of multimedia interactive installation at the Scarcrops Museums based on sensors, actuators and UDOO boards.
The Buonconvento sharecropping museum (Siena, IT) in 20xx was subject to a flood and all the multimedia installations were destroyed. The project restored the previous installations through the use of UDOO embedded platforms which allowed to exponentially reduce consumption and introduce new interactive possibilities based on the use of sensors and programmable by unqualified museum staff via the app inventor visual platform. .