Asset failure is an unavoidable part in every industrial process and thus it is necessary to count-on with the necessary tools that provide early warnings in case of potential or catastrophic failures. Notabily, predictive maintenance is central for systems and devices that are being used in the healthcare industry as well as refrigeration assets used in supermakets. Asset failure prediction uses data from sensing devices, that are continuously collecting the state of these systems and its surroundings, with the goal of providing better management, optimizing energy consumption and ensure environmental policy compliances.
The technology that allows this type of prediction relies on signal processing, sensor fusion and machine learning. Signal Processing is used to extract useful information from noisy signals while Sensor Fusion technologies are used with the purpose of reduce uncertainty on the measurements. Finally, Machine Learning and Artificial Intelligence tools predict the likelihood of failure within certain timeframe.

I collaborated in research involving human detection and counting in indoor environments (classrooms, offices). In this project, RGB-D cameras (Kinect and Intel RealSense) were used to capture depth information that later on was used to perform human detection and counting using convolutional neural networks. The occupancy information was then used to control heating and ventilation systems in order to improve building energy usage.

As a research associate, (under the guidance of CMU Professor, Javad Mohammadi), I collaborated on projects related to building energy management and retail analytics for one of the largets supermakets in the USA and partnerships industries in Thailand respectively. In particular, I worked on real-time analytics in order to increase inventory turnover, track product freshness and optimize warehouse through inventory management.