Oscar E. Mendoza

I earned my PhD with my dissertation titled "Adaptation of Neural-enhanced Retrieval Models to Domain-specific Tasks." My research focused on enhancing retrieval models for domain-specific applications, particularly in medical and academic search. By integrating advanced Transformers (autoencoders and sec2sec) models and natural language processing techniques, I developed methodologies that significantly improve the precision and relevance of search results in specialized fields

In my roles as a Data Scientist, I developed and maintained robust Python pipelines for educational data analysis and analysis of surveys. My responsibilities included designing test assessments, detecting cheating patterns, analyzing educational results, and performing audits. I also worked on database integration and developed NLP/ML pipelines for automated essay evaluation, delivering production-ready solutions to complex data challenges.

During my undergraduate and master's studies at the Industrial University of Santander, I focused on advanced research in image processing and optimization algorithms. My bachelor's thesis involved modeling and simulating a compressive sampling system for computed tomography (CT). This research demonstrated how compressive sampling could reduce the number of X-ray projections needed for high-quality 3D image reconstruction, using coded apertures to enhance image quality while minimizing radiation exposure.

In my master's program, I advanced this research by developing an algorithm for spectral image fusion in the compressed domain to improve spatio-spectral resolution. I formulated and solved an inverse problem using the Alternating Direction Method of Multipliers (ADMM) to optimize sensor parameters and enhance image reconstruction quality. This work aimed to integrate high-resolution spatial data with high-resolution spectral data, addressing challenges in remote sensing applications.