Miguel Angel Salinas Gancedo

Official researcher in the SIMUR Research Group (Multisensor systems and robotics group) for my Doctoral Thesis: Visual Analysis and Recommendation systems for big Datasets

Analysis and development a Machine Learning Pipeline to automatize the pretreatment, training and validation of machine learning models to analyze the health status of a cohort of 100 participants based on sensors: accelerometers, gyroscopes. This project is based on the study protocols proposed by expert groups for the validation of device-based outcomes for physical activity supported by cohort IMPACT

The the tech stack used is:

- Python: scikit learn, pandas, tensorflow, keras, etc

- Machine Learning: Convolution Neural Networks, Randforest.

- Formats: binaryy and csv files.

- Matplotlib for visualization.

My involvement in the support and development of the core framework of the company Consum includes:

- Analize and Design Backend:

- Analize and Design Frontend UX/UI

- Follow Agile Methodologies

- Technical documentation

- Techstack: Angular 19, Springboot 3.4, Openshift, Jenkins, SonarQube, Agile Methodology

1) Analysis, architecture, design, programming and deployment a scalable system capable of visualizing high-dimensional data in an agile way on the web, using Machine Learning techniques and standards frameworks.

2) The solution was initialize hybrid using Azure Private Repositories to manage deployment artefactors: docker images and kubernetes charts. We use Agile Project methodology to control de project using Azure Develops. Finally we deploye all deployment CICD stack on promise.

3) The datasets used for the project was: GDC Datasets from National Cancer Institude and Electricity consumption at the Leon hospital Datasets

2) The Tech Stack was

- Microservics architecture

- Backend Java SpringBoot microservices: JPA mongo, Docker SDK, etc

- Backend Python Slak microservices: skitlearn, t-SNE algoritm, pandas, bokeh, etc

- Frontend Angular services: PrimeNG to design de UI and WebGL (Regl) technics to use GPU to visualize more than 10000 points in a chart smoozly

- IAM Security Service based on Redhat Keycloak using OUth2 standard.

- Minio Object Storage for datasets scalability.

- Apache RabbitMQ Message Broker for realtime comunication

- MongoDB Business persistence.

- PostgreSQL for Security persistence by Keycloak

- Docker microservices.

- Kubernetes service orchestrator.

- Kubernetes Charts to be deployed.

- HAProxy Proxu service.

- CICD based on Jenkins and Nexus.