Carlos J. Gil Bellosta

I led a project to develop a large-scale prediction model for the demand for medical devices, which is crucial for optimizing their logistics and distribution. Leveraging machine learning techniques, I analyzed diverse data sources, including infection rates, healthcare capacities, and supply chain constraints. The model accurately forecasted the demand for these essential devices, enabling better resource allocation.

By integrating real-time data and predictive analytics, the model helped reduce shortages, minimize overstock, and ensure timely delivery of critical supplies to high-need areas. Collaboration with cross-functional teams, including healthcare professionals and supply chain experts, was key to refining the model's accuracy.

An official application sponsored by the Ministry of Health to track COVID-related mortality in Spain. I developed the statistical models behind the application, which are based on a GAM Poisson model that includes controls for heatwaves, cold spells, age, sex, and other covariates.

I created models to estimate openings, clicks, and conversions on email marketing campaigns at the customer level. Then I created a system to optimize the weekly number of sendings to minimize the number of contacts while maintaining the required profitability level of the individual campaigns.

A global consultancy was interested in building an HR dashboard to provide information on its hiring funnel, which I created and maintained. I created a database that integrated data from different onboarding systems, integrated them in a combined database, and created dynamic reports and dashboards for the global, distributed HR team.

I implemented several state-of-the-art, large-scale prediction algorithms on thousands of sales-related time series for a major supermarket chain. I assessed their performance in terms of accuracy and other business goals, such as robustness, training speed, and maintainability.

I handled a data-centric project involving sensor data from several health-related devices. The data was properly cleaned, and an embedding was created from the raw signal data. These embeddings were then stored in a vector database for retrieval and analysis. Finally, to visualize and analyze the results, I developed a custom dashboard tailored to the project's specific needs.

The customer is a large consultancy firm that specializes in eCommerce and marketplace strategies. I worked with them to create a model to optimize bids on online marketplaces for a variety of categories and products based on historical performance and using advanced SOTA time series prediction and Bayesian statistical methods.