Oussama Zaibi

• Developed synthetic image documents to fine-tune a Donut Transformer. • Conducted extensive preprocessing techniques for low-quality document images. • Finetunned OCR-free end-to-end Transformer model for image classification and invoice parsing achieving an 85% f1-score. • Addressed security and ethical implications of data extraction in forensics.

As a team member in an early-stage startup, I contributed to the development of a low-code, data-centric platform simplifying data management for data scientists. Specifically: • Tested and implemented new features to enhance platform functionality • Identified and resolved product issues to improve stability and user experience • Conducted research and developed pocs for image analysis features to expand platform capabilities • Developed and refined prompt engineering techniques to optimize LLM-generated SQL outputs • Collaborated on defining and implementing data contracts for dbt to ensure data quality and consistency

• Designed a remote temperature and humidity detection unit for servers room with the goal of improving response times to unexpected cooling system problems. My work resulted in a faster and more effective intervention process, helping to minimise downtime and prevent damage to critical systems.

• Analyzed and modeled Amazon’s product ranking patterns using time-series techniques to leverage insights for client advantage • Detected noise in scraped data and collaborated with data engineers to improve data quality and reliability