Exploring and building new data services for customers of the Ericsson cloud radio access network.Developing the Ericsson data acquisition and MLOps platform.

Collaborating with Data Scientists, Developers, UX Designers and Product Managers to help steer product and projects with data informed decisions.Full stack data scientist, from elaboration of schema for events reporting to building and maintaining ETL and models (network theory, clustering, recommendations) on top of pySpark infrastructure, making use of GraphX amongst other.Performing analysis through SQL (Presto/Trino), pySpark, Jupyter Notebooks and Python Data Scientist environment (pandas, numpy, scikit-learn, scipy, etc.).Data story telling, KPI elaboration and dashboard creation (Mode Analytics) to communicate the insights.

Being part of a small team of Data Scientists / Machine Learning Scientists I came to:-Built an automatic fault detection model which fasten detection and correction of faulty telecommunication nodes and thus improves in-service performance and lowers failure costs and failures analysis cost. It is based on a One-shot learning approach through Siamese Deep Neural Network (Keras/Tensorflow).-Built a classifier to identify IoT devices from Human users and thus allowing differential charging based on the class of users. It is based on an LSTM DNN (Keras/Tensorflow).-Built a mobile device subscribers behaviour analysis which allows telecommunication operators to better segment and serve their customers. It is based mainly on clustering and regression (scikit-learn, pandas, numpy, ...)-Developed the Architecture for a distributed data pipeline and a model serving system for the telecommunication industry enabling Ericsson to deploy Machine Leaning models commercially.-Devised a way to perform subscriber survey analysis and gain automatic insight on their pain points, thus saving survey analysis time and allowing faster time to correct expressed issues. Is it based on Natural Language Processing (NLP) and sentiment analysis (senti-wordnet).-Built a predictor for support ticket SLA breach before they happen thus saving penalty costs (scikit-learn, pandas, numpy, ...).-Participated in various others IoT related analysis: car usage patterns and electrical meter fraud detection. This is based on clustering and outlier analysis (scikit-learn, pandas, numpy, ...).-Experimented with Text Generative Recurrent Neural Networks (RNN), Sentiment Analysis and Image Neural Style Transfer (Keras/Tensorflow).

Leading a small research team, we created a Polyglot Hybrid/Multi-Cloud Application Development Framework based on the Actor Model and applied to the Telecommunication Domain and to the IoT. Refer to my Master thesis for details.Also did some research in analytic model to predict critical care patient pain level based on vital signs, using Machine Learning algorithms and Genetic Algorithms (Matlab).

Leading a small research team and responsible for Research Coordination in collaboration with Universities on Software Clones detection and implementation in industry as a developer toolbox.Principal researcher for Domain Specific Language development and research initiative applied to IMS communications.Research on Genetic Programming.

Software Architect, Innovation Team Leader and Developer for a small footprint complete telephony system for disaster relief / public safety applications.Software Architect, Innovation Team Leader and Developer for various Ericsson converged SIP/HTTP Applications including elements of social networking and user application development.Software Architect, Innovation Team Leader and Developer for a Full Stack Community Collaborative Ideation Tool development for Ericsson internal users.

Technical Lead and Software Architect for a team of up to 32 persons developing real-time software for Ericsson CDMA Base Station Controller (MAC/SOE functionality).

Real-time software development for Ericsson CDMA Base Station Controller, Ericsson TDMA Radio Base Stations and Ericsson AXE switching center.

Performed CF-18 real-time Software Support. Preliminary study, impact study and implementation detailed plan preparation for implementation of a new bomb-release algorithm.Development, test and installation of real-time virtual battlefield software (ITEMS) that can be used stand alone or as part of dedicated military simulation equipment. Specialising in entities position keeping model (flat, spherical and elliptical earth models); missiles, rockets, bullets, torpedoes and depth charges dynamic model (6 d.o.f.)