In July 2013 I founded Mivoq SRL with three of my former colleagues and one entrepreneur, with the goal to let everybody to create their own synthetic voice. In 2012 we presented the idea at StartCupVeneto business plan competition and we were awarded the first prize. We also won a few other competitions and we were able to get management training and some funds to start the company and finally in 2013 we started it.Working for a small startup with very limited budget has been very challenging, but it has been also a great experience that let me grow a lot.I have been the CTO for Mivoq until January 2016, when, after a very intense period, I decided to leave that position in order to alleviate my workload, at least in terms of formal responsibilities, by leaving them to the advisory board, of which I am a member.Being Mivoq a very small company, I have been responsible for almost any technical aspects, from software to hardware, from design to implementation, from candidate interviews to project management, from technology reviews to IT administration and I actively contributed in all the aspects of the company.

I rejoined CNR to work on ALIZ-E European project, whose goal was to develop a system for child-robot interaction, based on Nao robot.My team was responsible of development of Automatic Speech Recognition and Text to Speech modules and of their integration in the core system. I was responsible of developing Natural Language Processing modules for the TTS system, Language and Acustic Models for the ASR, forced alignment systems, corpora development and managment (including the definition of annotation schemes and support tools if needed). My duties also included writing scientific papers and technical reports, partecipating to conferences and project meetings, collecting data during experiments and recording sessions, coordinating and supervising annotators.

As a team lead I have been managing a team of 30+ people working on Bixby voice assistant by Samsung. The team included linguists, developers and UX designers.We worked in an international environment, interacting with multiple stakeholders in different timezones.

Working at IRIS unit at San Raffaele has been a wonderful experience. I had the chance to work with many skilled and motivated people. I really appreciated the close interaction with the design unit.I was hired to support the final phase of TAPPS European project. I was responsible to technically lead a small development team. My daily duties included writing/coordinating deliverables, implementation of core components in real-time, multi process environment, planning development activities, support the PM.

Working at CINECA was amazing and I managed to improve all my skills there. I was hired to work on information extraction from multimedia content in the context of a European project. I had the responsibility to develop applications and prototypes for:- Speaker Diarisation (including Speaker clustering, Gender classification, Speech/Non-speech classification and, later, Speaker Identification). Among the most important results were: smooth integration/adaptation of existing opensource technologies, a novel algorithm for Speaker clustering based on Genetic Algorithms and Mutual Information, a pitch extraction algorithm and the creation of data for all the above applications. I also developed a few web demos and a small video player capable of showing the results of these analysis applications;- Document Summarisation (including Topic classification, Named Entity Recognition, Shallow Parsing, Keyword extraction, Ontology Concept linking, Ontology-based search engine). Among the most important results were the smooth integration and adaptation of many opensource technologies, the addition of multilanguage support to many of them (the project required English, French, German and CINECA was interested in Italian), the creation of data for most of these applications. Later I also developed several demo applications based on these technologies and collaborated with other working groups in order to propose the adoption of these technologies (e.g., apply Ontology-based search to recorded audio/video university lessons);- Speech recognition: my work was about integration of existing speech recognizer, data development (e.g., Language Models creation) and optimization of the recognizer for the specific task. Among the most important results were a testing framework that allowed to easily propose setup changes and comparison with other setups: this framework included the adaptation of existing evaluation tools to the specific task and the definition of appropriate methods;

I worked at Expert System for a very small time in a group developing Automatic Speech Recognition. The engine was written in C, with a Tomcat+Javascript job scheduler interface. My main duties were to debug the ASR engine and parallelize it. I also implemented a simple energy-based voice activity detector in C. I started prototyping a job scheduler as a Tomcat webservice, with a Javascript-based frontend (using qooxdoo).

I started collaborating with Piero Cosi at CNR during my master thesis. I mainly worked on engineering a 3D facial animation engine, Lucia. I was able to split the engine into a library and a few frontend applications (I also introduced a new frontend based on Gtk2), to reduce the redundancy in its code (the final codebase was 1/4 of the original one), to increase its speed (4x, on average, thanks to better memory allocation), to create a better eye model, to port it to Linux. I also worked on automatic text to video synthesis software AVengine, porting it to Linux and on speech synthesis software (Festival, SMS, Mbrola, OGI diphone engine), updating and maintaining them. I was able to learn a lot about developing and maintaining multiplatform codebase and how to automatize configuration and building.I also introduced in the laboratory the use of Linux, Virtual Machines for web servers (kvm), Virtual Private Network (OpenVPN), Version Control (Subversion), internal DNS and DHCP, by creating initial configuration of all these services, writing scripts and configuring complementary services (e.g., iptables, failtoban, cron scripts for backup, ...). Most of these services are still in use today.

I worked at WinVaria for a very limited time. My main responsibility was to start the migration of their internal database to Postgres and to prepare Joomla plugins to allow access to the data. I also developed a small C# application for a barcode scanner (Windows CE) to synchronize with their main software. I also setup a Linux box with iptables to allow portmapping (as a temporary solution due to hardware failure).