Victor Hernandez

Victor is working with colleagues from B3 in Örebro and Sundsvall on a project that uses A.I. in the educational industry. Victor and his colleagues implement algorithms for recommending syllabi and didactic materials to teachers and education administrators from different schools. The project is about machine learning, NLP, and unsupervised learning. The project is being developed if the Azure ML platform and solution are to be a complete A.I. pipeline.

Victor worked on a project relatd to ePsychiatry and the prediction of re-enrollments in addiction care. To predict re-enrollment, it is required a system that can use structured data and unstructured data. Structured data consists of patient's data, such as age, gender, marital status and several different data in visit history. Unstructured data refers of notes from doctors and healthcare professionals. The goal of the project was to use both structured and unstructured data to predict if a patient would return within 15-30 days. Victor was part of a team of four consultants and he worked with several types of machine learning algorithms and techniques, such as natural language processing, deep and shallow learning, data analysis and algorithm optimization.

Victor proposed technical solutions to two different problems in the retail industry. The analysis was performed for a customer that provides forecasting services to retail companies. For the first part of the analysis, Victor identified shortcommings of the solution that is currently currently used to predict customer engagement in the form of webpage visits. Victor suggested the use of multivariate machine learning algorithms to make forecasts in the form of confidence intervals instead of point prediction. In the second part of the analysis, Victor identified a solution that uses "reinforcement learning" for smart ad placement.

Victor successfully developed an AI system for the Anicura group, which is a network of veterinary clinics across Europe. Anicura aimed to harmonize patient record systems. AniCura's goal was to go from over 40 systems to only 2. The challenge is to harmonize and migrate free text data from all the different systems to the two new ones. Among other things, a master list of offered veterinary services would be created.

NLP is a suitable approach for finding clinics that offer similar services. Victor led the work where different NLP strategies were implemented and tested. The result was a system that could perform data harmonization not only based on free text, but also be able to use additional metadata. Victor was responsible for presenting results and insight to the customer in such a way that strengths and opportunities for improvement could be easily identified and understood.

Victor oversaw the work of students for the 2019 bachelor's degree thesis - related to the prediction of hip dysplasia using artificial intelligence. In addition, during this time Victor gave several lectures in various places, such as TechDays and workshops with internal seminars.

Victor implemented a prototype to perform "novelty detection" in shipping containers. The solution detected tip overs of heavy goods during transportation. This problem can become an occupational risk when the cargo reach its final destination and human operators perform unloading operations. The models were trained with time series from sensors mounted on the transport vehicles.

Victor developed a PoC that used NLP to reduce testing time in the production line of a telecommunications company. The solution aimed at finding simiar test by matching free text that described the testing procedures.The plan was to use the solution along with additional input from the experts to discard the tests that are redundant.

Victor implemented a PoC for Örebro's largest housing company. The PoC was about predictions of electricity consumption in households. Victor took a large volume of time series data from several buildings and created a model that predicted electricity consumption of existing and planned buildings. The model estimated the importance of factors that affect electricity consumption. In addition, the model predicted breakpointswhen consumption patterns suddenly changed or showed abnormal behaviour.

Victor worked as a postdoctoral researcher in machine learning for 5 years. He worked in several EU and national projects. He collaborated with several industry partners as well as research institutes such as the German Aerospace Center (DLR), Tokyo University of Agriculture and Technology, The University of Warwick and the Fluid Mechanics Laboratory at Cornell University. Victor developed several technical solutions that combined sensor data, robots and machine learning algorithms to create statistical models and to predict different variables such as gas distribution, air quality and air flow. In 2019, Victor applied for and received funding for his own research project. He received two international awards and published more than 50 articles in several scientific journals.

Victor was awarded a Ph.D. in computer science in 2015. His research topic focused on machine learning applied to environmental monitoring robots. In such area of research, Victor developed different statistical models for tasks such as prediction of pollution levels. This task requires, for example, the implementation and quantitative/qualitative analysis of different regression models and the selection of informative/useful variables. In addition, Victor worked with different machine learning techniques for classification, unsupervised learning and optimization.

Victor worked in projects related to Information and Entertainment radios for General Motors’ vehicles. Victor’s task included the implementing new functionality, conducting changes and testing.

Worked as a systems engineer for different HVAC projects for customers such as General Motors and Ford. As a systems engineer, Victor tasks include requirement management and definition, visiting the customer’s production site for in-vehicle testing and coordinating between the different competencies (e.g. electric, software and manufacturing engineering) during the project’s life cycle.

Victor developed and executed test procedures for embedded modules. In addition, he developed an automated testing tool for network communications. The tool allowed to perform functional test of network communications (e.g. CAN, LIN, J1850) and produced pass/fail reports. The tool was quickly adopted by several projects in Delphi.