Pavlos Aquino -Ellul

Worked on enterprise AI use cases focused on advanced document processing, including digitalization, segmentation, and intelligent information extraction. Led Generative AI and LLM-based solutions, designing and implementing systems for knowledge access and automation. Built conversational AI platforms for the Oil & Gas industry for a large Saudi Arabia organization, enabling natural-language interaction with large-scale technical and operational data.

Led end-to-end development of a high-performance OCR system for Greek historical documents, covering architecture, data collection, preprocessing, model training, and deployment. Designed and trained object detection and OCR models optimized for accuracy and scalability. Built and deployed Python-based OCR pipelines for large-scale production with continuous evaluation and improvement. Acted as technical lead for the OCR stack, driving research, prototyping, planning, and prioritization, and making key architectural and technical decisions.

Led end-to-end development of an e-Auscultation AI use case as part of a prototype-to-product mobile health initiative. Designed and trained machine learning models to detect wheezes and crackles from audio streams, including dataset preparation and experimentation with 1D/2D time-series approaches. Deployed models on-device using TinyML with Kotlin and TensorFlow Lite constraints, and integrated e-Auscultation and cough detection into a single smartphone application with user-controlled start/stop. Built a backend API to securely collect inference results and store them in a database, with privacy-by-design (audio processed locally; only results transmitted).