Ayoub sahlaoui

I’m a data scientist and machine-learning researcher with a foundation in software engineering. My experience spans
applied AI research, data-driven experimentation, and full-stack development.

In 2026, I completed a six-month Erasmus+ research internship at the University of Limerick, where I studied
parameter-efficient adaptation of V-JEPA 2 for egocentric action anticipation. Using LoRA, I adapted a 307-million-
parameter video model while training only 0.13% of its parameters. The best configuration improved action mean-class
Recall@5 from 16.39% to 18.67%, a 13.9% relative improvement. This research was accepted as a paper at IMVIP 2026,
and I defended my master’s thesis in June 2026.

Before moving deeper into data science and machine learning, I gained practical software-engineering experience at
Octobot Consulting. There, I contributed to an insurance management application built with Laravel, Angular, REST
APIs, JWT authentication, and MySQL.

My path through software engineering, data science, and academic research has given me a broad perspective on how
ideas become working systems. I’m especially interested in efficient machine learning, computer vision, and the
engineering behind dependable AI. I enjoy challenging problems, careful experimentation, and turning complex
technical work into conclusions that can be clearly understood.