Abderrahmene Hamdi

Abderrahmene Hamdi
Abderrahmene Hamdi
Born
🏫 Education	Nagoya University; Ecole nationale Supérieure d'Informatique; Higher School of Computer Science (ESI-SBA)
💼 Occupation	AI engineer, researcher
📆 Years active	2023–present
Known for	Work in machine learning, natural language processing, large language models, vision-language models, and health misinformation detection
🏅 Awards	Best Year Project Award (2024)

Abderrahmene Hamdi is an AI engineer and researcher whose work focuses on machine learning, deep learning, natural language processing, large language models, and vision-language models. His work has included research on multimodal AI, no-reference image quality assessment, health misinformation detection, retrieval-augmented generation, and GPU-accelerated model training.

He has studied and conducted research in Algeria, France, and Japan, and has worked on projects spanning language models, multimodal learning, distributed training, and AI engineering systems.

Early life and education

Hamdi studied computer science in Algeria. He completed a preparatory cycle in computer science at the Higher School of Computer Science (ESI-SBA) in Sidi Bel Abbès between 2020 and 2022.

He later attended the Ecole nationale Supérieure d'Informatique (ESI) in Algiers, where he completed an engineering degree in computer science in 2025. His final year emphasized applied research and included work on vision-language models for no-reference image quality assessment under joint supervision involving Sorbonne University.

In 2025, Hamdi joined Nagoya University as a MEXT Research Student in the Ryohei Sasano Laboratory. His research there focused on natural language processing and computational linguistics. He subsequently entered the university's master's program in Intelligent Systems, funded by the Japanese Government's MEXT Scholarship, with a focus on artificial intelligence, NLP, and large language models.

Career

HeReFaNMi / NGI-Search

From December 2023 to February 2025, Hamdi worked remotely as an AI / LLM Intern with HeReFaNMi, a project associated with NGI-Search. In this role, he contributed to the development of an end-to-end large language model pipeline for health-related fake news mitigation.

His work included data collection, annotation strategy, retrieval-augmented generation workflows, fine-tuning and customization of language models, prompt design, safety filtering, and automated evaluation using metrics such as precision, recall, and F1 score. He also worked on Docker packaging, CI/CD pipelines, deployment workflows, A/B testing support, model monitoring, and user-facing technical documentation.

Sorbonne Université

From January to July 2025, Hamdi worked as a Vision-Language Model Engineering Intern at Sorbonne University in Paris.

During this internship, he conducted literature reviews on contemporary vision-language models and designed experiments involving compression, medical imaging, and retrieval-augmented generation. He explored limitations of current VLM systems for no-reference image quality assessment and integrated perceptual components such as CLIP and SAM into evaluation workflows.

He also developed prompt-learning and prompt-tuning strategies, created multimodal data curation and augmentation pipelines, and implemented training scripts for AMD MI300X hardware using ROCm/HIP while supporting Nvidia GPU workflows through Slurm-based benchmarking. His work additionally included downstream evaluation tasks in medical image classification, captioning, and visual question answering.

Research and technical work

Hamdi's technical work spans machine learning, deep learning, natural language processing, computer vision, algorithm design, software engineering, information systems, and hardware optimization.

Among his listed projects is the full training of a 2.1 billion-parameter large language model from scratch using eight AMD MI300X GPUs. He has also re-implemented the GPT-2 and LLaMA 2 architectures in PyTorch.

Other projects associated with his portfolio include:

100 days of building GPU kernels – a public engineering log documenting daily GPU kernel development over 100 consecutive days.
Fennetic – a platform designed to guide visitors in Algeria using artificial intelligence.
Python-CUDA Transformer – CUDA-accelerated Transformer mathematical operations implemented primarily in Python.
ChaosVAE – a variational autoencoder-based model intended to learn and generate chaotic systems.
Illness Detection and Classification – a machine learning project for estimating likely illnesses from symptom sets.
Tasksly – a smart to-do application using AI and blockchain-related ideas.

Areas of expertise

Hamdi's listed areas of expertise include:

Machine learning
Deep learning
Large language models
Natural language processing
Data analysis
Information systems management
Human-computer interaction
Project management
Enterprise resource planning systems
Software development
Computer vision
Algorithm design and optimization
Hardware optimization

His technical stack includes programming and engineering work in Python, C++, Java, and R, as well as experience with CUDA, Triton, HIP, OpenCL, ROCm, Kubernetes, MPI, TensorFlow, PyTorch Lightning, MLflow, Apache Spark, Docker, AWS SageMaker, Google AI Platform, SQL, and LangChain.

Conference presentations

In November 2024, Hamdi presented work in Bolzano, Italy, under the title Health Misinformation Detection: A Chunking Strategy Integrated to Retrieval-Augmented Generation. The presentation was delivered at the Second Workshop on AI for Perception and Artificial Consciousness.

Workshops, outreach, and volunteering

Hamdi has participated in STEM outreach and AI education activities.

In 2023, he served as a STEM trainer, teaching university students linear algebra concepts including vectors, matrices, eigenvalues, and matrix operations, while relating them to machine learning applications.

He has also created artificial intelligence challenges for community events including the CSE Datathon and GDG DevFest, led a workshop on convolutional neural networks, and delivered a talk at a Google I/O Extended event in Algiers focused on recent developments in AI.

Awards and honors

In 2024, Hamdi received the Best Year Project Award from the Higher School of Computer Science (ESI). The award recognized project performance and leadership, including work on the development of a new university website. He also received a certificate of honor presented by the university director.

Certifications

Hamdi's listed online courses and certifications include:

Prompt Engineering for ChatGPT (Vanderbilt University / Coursera, 2023)
Data Analyst Associate Certificate (DataCamp, 2023)
Supervised Machine Learning: Regression and Classification (Coursera, 2022)
Apply Generative Adversarial Networks (GANs) (Coursera, 2022)
Sequence Models (Coursera, 2022)
Convolutional Neural Networks (Coursera, 2022)

Languages

Hamdi's listed languages are:

Arabic – native
English – full professional proficiency
French – professional working proficiency
Japanese – professional working proficiency (N1)
Chinese – basic proficiency

Selected reading

A list of technical readings associated with Hamdi's profile includes:

Pattern Recognition and Machine Learning by Christopher Bishop
Build a Large Language Model (From Scratch) by Sebastian Raschka
Machine Learning: From Theory to Algorithms by Shai Ben-David and Shai Shalev-Shwartz
Convex Optimization by Stephen Boyd
Introduction to Statistics for Data Analysis by David M. Diez
PyTorch Natural Language Processing Programming: Analyzing Japanese Text with Word2Vec, LSTM, Seq2Seq, and BERT by Hiroyuki Shinnou
Calculus on Manifolds by Michael Spivak

External links

This article "Hamdi Abderrahmene" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:Hamdi Abderrahmene. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.