Profile
I am an AI/ML Engineer working at the intersection of applied research, experimentation, and real-world machine learning systems.
I’m interested in turning research ideas into reliable systems. I work across a broad range of machine learning topics, with particular depth in computer vision, including visual recognition, OCR, pose estimation, segmentation, and video understanding.
I tend to start from the problem rather than from a specific model: what data is available, what constraints matter, how the output will be used, and what combination of techniques can produce a practical result.
Computer vision is my strongest area, but not my only lens.
At L’Oréal, I developed computer vision and data science solutions for product evaluation, laboratory automation, and robotic applications. This work involved low-data settings, synthetic data, experimental design, segmentation, object detection, and deployment under limited computing resources.
At Djezzy, I designed an end-to-end OCR pipeline for extracting and validating information from biometric identity documents captured under uncontrolled conditions.
At CDTA, I worked on 3D object pose estimation for a medical augmented-reality project, studying different pose-estimation approaches, neural architectures, and evaluation benchmarks.
At Hitachi Rail, I work on detecting passenger medical-distress situations in dense railway environments, combining person detection, tracking, pose estimation, and behavioral analysis under edge-deployment constraints.
I usually begin by studying the problem domain and reviewing the state of the art. I read research papers, examine existing methods and benchmarks, compare their assumptions, and identify their limitations before deciding what is worth reproducing, adapting, or challenging.
From there, I establish clear baselines and move through experimentation: testing assumptions, studying failure cases, measuring results, and iterating until I understand not only whether a method works, but why.
I care about reproducibility, interpretability, and deployment constraints. A model is only one part of the solution: data quality, evaluation protocols, maintainability, computing resources, and the people using the system are equally important.
Beyond computer vision, I’m interested in generative AI, retrieval-augmented generation, NLP, multimodal systems, and intelligent agents. I’m especially interested in problems where these techniques can complement perception models—for example, by adding contextual reasoning, knowledge retrieval, or natural-language interaction to visual systems.
Pride and Prejudice and 1984.
RPGs, especially The Witcher.
Manga across different genres.
Princess Mononoke (Mononoke Hime).
Native
Professional proficiency