Investigated characteristics of human mobility and their impact on the network data traffic, planning and deployment. I've analyzed large-scale datasets from mobility and traffic demands generated by millions of users. Python's multiprocessing, thread and R multi-core libraries were used to summarize and assess massive amounts of data.

Among the results of this work there are 5 published conference papers, 2 published journal papers and a synthetic data traffic simulator.

Studied the usage of centrality information on the design of routing algorithms for Wireless Sensor Networks. I have proposed new topological metrics, distributed algorithms to calculate them, and a tree-based routing algorithms that take advantage of those metrics.

Among the results of this work there are 1 published conference papers, 1 published journal paper.

Created a wireless sensor's application for fire detection using real dataset of weather measurements. Proposed a novel way to fuse and codify data.

Among the results of this work, I have published 1 conference paper.

Developing features, optimizing performance, deploying to production and monitoring a dozen of Java services at the Trip Search team. For example, meeting points service to provide drivers and passengers where to meet using Machine Learning recommendation. Routing service based on OSRM to fastly calculate routes at user search time to match drivers and passengers with the smallest possible detour. Mainly technologies are Java, XGBoost, R, Python, OSRM, PostgreSQL, PostGIS, MySQL, Elasticsearch, Redis.

Developed machine-learning assisted techniques to predict users' QoE based on network's KPIs. Among the results of this work, 1 conference paper has been accepted for publication.

Developed in Ruby on Rails, as part of a team, Telecom Italia Mobile's project TIM Beta

Developed, in Python, a project called Future of Enterprises aiming to improve the real-time interaction among employees.

Lead developer of a social network in Ruby on Rails.

It provides a full-featured face recognition SaaS capable of predicting age, emotion, ethnicity, gender, landmark, glasses, smile, hair for all the faces on an image. All the predictions are made using end-to-end machine learning models written in Keras and Tensorflow. Besides it calculates head-pose, face bounds, exposure and blur level. A fully documented API is available here.

Eazy.bike finds the best bike-sharing cycling routes in more than 443 cities distributed in 45 countries. It uses real-time information of bike availability and proximity to predict the best stations to pick-up and drop-off. The Android and iOS apps provide automatic trajectory redirection whenever destination station fills up during user's ride. It can be accessed by city-specific domains for improved context information, e.g., paris.eazy.bike, or london.eazy.bike, etc.

Collaborative application that aims to publicly open the Brazilian Postal Code (CEP) data. Contains information of about 1 million CEPs. It provides an API for developers and, for the end-users, a collaborative platform to improve the quality of the data. About 30K registered users.

It helps people to choose companies based on problems presented with other customers. Received Honorary Mention award from the Brazilian Ministry of Justice.

Those applets improve Cairo-dock experience by enabling consistent usage of twitter, translation, search, moon phases, quote of the day, etc.

The "Repeat one song" feature for Rhythmbox.

Open Data Applications Contest. Lead developer of the Proconfie application.

Second best overall score during the bachelor.

For the work "Game Theory for Decision Making in Wireless Sensor Networks" presented in the 15^o Sciences Seminar.