I am or was principal researcher and technical lead in the following R&D projects and project manager for most of them.


2016: Chief Scientist for LiteMe: Nonintrusive appliance load monitoring using machine learning to disaggregate electrical signals from an aggregate signal. Currently under development. Uses an ensemble of classification methods on time series.

Results: Under development.

Electoral Radar

2014: Data analysis system for electoral campaigns. Obtains data from social networks and other public sources and uses machine learning algorithms and a business intelligence interface to help user to gain information and insights.

Results: Used by 5 political campaigns in 2014. See https://jpsauve.net/2015/11/19/electoralradar/

Smart Alarms

2004-2008. Software system to help power system operators diagnose system faults during large scale contingencies.

Results: Used in 6 control centers in a major Brazilian power generation and transmission utility. Very successful in providing diagnostics in real-time. Currently (2016) being evaluated by other Brazilian utilities.


EasyAccept is a tool that helps you create and run acceptance tests in an easy, quick, and clean way, bridging the communication gap between clients and software developers.
Bindings available for Java and Python. I conceptualized the tool and coded the first version for Java.

Results: Available on github; used by several local companies.

Bottom Line

2004-2009: Business-Driven Information Technology Management (BDIM). Development of models to bridge the business-IT gap and allow IT to be managed with an eye on business impact. The project focused on change management, problem management, performance management, and portfolio management.

Results: several academic papers were published and a best paper award received.

Smart Action

The Smart Action project aims to develop a software system to help power system operators perform System Restoration at a very high level.

Results: the system was fully developed and tested but proved to incur too a large configuration debt to be used in production.

Smart Analysis

2008-2010: Project aimed to specify, implement and test a software system to produce diagnostics of faults in power systems equipment and systems. The software system is based on behavior models of such equipments and systems.

Results: Project successfully concluded. Never entered production due to lack of engineers to develop behavior models.

Smart Switch

2010. Software system to help power system operators find optimal switching  sequences to isolate or restore parts of a power system while obeying interlocking and other restrictions. An automatic solution lowers restoration times and minimizes errors.

Results: Used in 6 control centers in a major Brazilian power generation and transmission utility.