Redesigned and reprogrammed the business core code of the company, which was implemented through conditional programming on C# to inference models based on Machine Learning using Python as the principal tool. Tensorflow, Pycaret, Keras and Scikit-learn are some of the applied libraries.Implemented an adaptative algorithm for algorithm parameters based on non-supervised model KMeans to avoid the work overload of the support team as the parameters were manually modified.

✔ Implemented a routine to filter and amend odometry data from vehicles of transport and mining companies. The project aimed to eliminate specific errors produced by hardware and hence increment detection rate. The project had three stages: • Data Analysis: Built visualization tools for data errors to separate them from the groundtruth. Moreover, setting up the different sets for training and testing. • Criteria Design: Designed the metrics and criteria to match the detection and the Ground Truth, avoiding sensitivity and specificity errors. • Prototype Design: Handled and Stored the data in a custom database; this included extraction, validation, estimation, trip detection, and data storage.✔ Developed a routine to validate and filter raw data. The raw data possessed errors from the firmware and the data gathering process from the vehicle. The technology stack used: Python, Postgresql, and Google Cloud.

Remote Work. Programmed software routines focused on pattern detection based on distance, current, RFID, accelerometer sensors. The aim of the project was the mine production control in Chilean and Peruvian companies.

• Lead backend software development for control production in mining companies. I developed the LHD (Load, Haul and Dump), Truck and Train production control based on RFID and inductive sensors}• Programmed sensor failure detection algorithms in C#. Applied machine learning-focused paradigms to solve detection problems• Designed and implemented a tracking algorithm utilising wireless sensors to show the path of trucks on a web interface.• Implemented a train production control system. The programming of this routine was based on a server-centric algorithm logic.• Modelled data structures through SQL database and the Microsoft SQL Server engine, applying Language Integrated Query tools.• Led the software team to develop an anti-collision and traffic control system. The purpose was to minimize the number of accidents in a high-risk zone, allowing the vehicles to work coordinately (tandem) and developing a communication protocol based on server-client architecture over the Node JS environment platform.• Redesigned and ported to C++ the LHD's production control algorithm for the software packing. The system achieved a higher performance; 80% less memory usage and lower execution time.• Designed and developed a packing solution through a unified model for all the signal analysis algorithms. This aiming to obtain a re-utilizable, maintainable, and scalable code with a good performance. I received the Mining Tag S.A. innovation award.

Partnership between Universidad de Chile and Neurologist Department of Clinical Hospital from Universidad Católica de Chile. Developing a routine with the aim of seizure detection in real time, over electroencephalogram records of eighteen channels with longitudinal configuration. Databases given by Neurologist Department, with focus in critical patients. Program algorithms in Matlab. Seizure labeling was made with the expert, then the set result divided in train and test, and the last was contrasted with the system outputs. The results was over 85% of success rate. The research was applied to “X Congreso Latinoamericano de Epilepsia”as a poster and was awarded.

Analysis and solving of statics programing issues, with the aim of cleaning and optimize code and remove “Design Compiler” software setbacks. Programing in C and C++.It was able solve 140 code bugs.