Welcome! I'm Kamel Guerda.
An AI Research Engineer at IDRIS – CNRS, advancing AI research and education on the French supercomputer Jean Zay.
Expertise
Data Analysis
Extract meaningful patterns and key insights to support smart, strategic decisions.
Model Design & Development
Design, test, and refine ML, DL, and LLM solutions tailored to complex challenges.
Optimize & Scale
Benchmark, optimize, and distribute models for fast, scalable training on GPUs and clusters.
Share, Teach & Transfer
Empower teams through tailored trainings, clear content, and sustainable solutions they can own.
Skills
Concepts
Programming Languages
AI Libraries & Frameworks
Infrastructure & Orchestration
Tools & Productivity
Languages
Talks
Monthly podcast on AI news, papers & innovations (YouTube)
Unlock the power of Supercomputing
AI & Cybersecurity
AI infrastructures & academic ecosystem
Deep Learning: Optimization & Acceleration
Frugal & Sustainable AI
Data Hell: Too little, wrong, or exotic data
Introduction to Deep Learning
Courses
A hands-on course to discover the fundamentals of deep learning and gain first practical experience with PyTorch. It is also an excellent introduction to understand the key challenges, stakes, and applications of deep learning, including for non-technical or decision-making profiles.
- Neural networks: context, definitions, fundamentals
- Practical exercises with graphical tools and PyTorch
- Methodology: from data management to model training
- Hands-on sessions on the Jean Zay supercomputer
A course dedicated to exploring the main deep learning architectures for a variety of data types and challenges.
- Convolutional Neural Networks (CNN)
- Recurrent Neural Networks (RNN)
- Transformers
- Graph Neural Networks (GNN)
- Diffusion models
- Hands-on sessions on the Jean Zay supercomputer
Advanced course on optimization techniques and large-scale, multi-GPU training for deep learning on a supercomputer.
- GPU acceleration and mixed precision training
- Distributed training and data parallelism
- Data processing and loading optimization, storage management
- Large batch training techniques
- Hyperparameter optimization (HPO)
- JIT compilation
- Large model distribution (e.g. LLMs) : Tensor, pipeline, ... parallelisms, ZeRO, FSDP,...
- Best practices and visualization tools (TensorBoard, MLflow, W&B)
A course focused on both the theoretical foundations and practical implementation of Large Language Models (LLMs), covering fine-tuning, prompt engineering, and deployment.
- Transformer fundamentals and language modeling
- Classical fine-tuning and PEFT methods (LoRA, RAG, Chain of Thought...)
- Setting up LLM training and optimization environments
- Data cleaning, evaluation, hyperparameter tuning, and production deployment
- Utilization of the Jean Zay supercomputer
Career
AI Research Engineer
@ IDRIS (CNRS) – Jean Zay Supercomputer HostApr 2021 — Present, Orsay, France (hybrid)
Advancing AI research and education on the Jean Zay supercomputer, with a focus on distributed training, interdisciplinary projects, and public outreach.
►Led the MINERVA Service work package in the European HPC-AI consortium to enhance HPC-AI integration.
►Contributed to interdisciplinary AI projects under the PNRIA framework and collaborated on the PLEAIS project, assisting in the training and deployment of a French LLM.
►Specialized in distributed training and optimization of large-scale AI models across multi-GPU and multi-node environments.
►Benchmarked JIT solutions and supervised an intern focused on performance analysis and distributed workflows.
►Developed and led training programs (IPDL, ArchDL, DLO-JZ, FIDLE) and mentored researchers during NVIDIA-sponsored hackathons.
►Engaged in public outreach through Fête de la Science and the Panoram'IA YouTube series to popularize AI concepts.
Apr 2021 — Present
Orsay, France (hybrid)
AI Research Engineer
@ IDRIS (CNRS) – Jean Zay Supercomputer HostAdvancing AI research and education on the Jean Zay supercomputer, with a focus on distributed training, interdisciplinary projects, and public outreach.
►Led the MINERVA Service work package in the European HPC-AI consortium to enhance HPC-AI integration.
►Contributed to interdisciplinary AI projects under the PNRIA framework and collaborated on the PLEAIS project, assisting in the training and deployment of a French LLM.
►Specialized in distributed training and optimization of large-scale AI models across multi-GPU and multi-node environments.
►Benchmarked JIT solutions and supervised an intern focused on performance analysis and distributed workflows.
►Developed and led training programs (IPDL, ArchDL, DLO-JZ, FIDLE) and mentored researchers during NVIDIA-sponsored hackathons.
►Engaged in public outreach through Fête de la Science and the Panoram'IA YouTube series to popularize AI concepts.
Technology Teacher
@ Académie de Paris – French School SystemJan 2021 — Apr 2021, Paris, France (on-site)
Taught technology and introduced artificial intelligence concepts to middle school students from Montaigne and Lavoisier schools through interactive, hands-on projects.
►Introduced students to AI concepts, including simplified machine learning and deep learning using Scratch.
►Guided 3D modeling and 3D printing projects, supporting students in transforming digital designs into physical prototypes.
►Taught the fundamentals of formulating and understanding technical specifications.
►Promoted curiosity, creativity, and engagement in STEM fields through interactive learning approaches.
Technology Teacher
@ Académie de Paris – French School SystemTaught technology and introduced artificial intelligence concepts to middle school students from Montaigne and Lavoisier schools through interactive, hands-on projects.
►Introduced students to AI concepts, including simplified machine learning and deep learning using Scratch.
►Guided 3D modeling and 3D printing projects, supporting students in transforming digital designs into physical prototypes.
►Taught the fundamentals of formulating and understanding technical specifications.
►Promoted curiosity, creativity, and engagement in STEM fields through interactive learning approaches.
Jan 2021 — Apr 2021
Paris, France (on-site)
AI Engineer
@ COSE – Aerial imaging systemsMar 2019 — Dec 2019, Paris, France (on-site)
Developed innovative computer vision and deep learning solutions for aerial image analysis, from research to real-time deployment.
►Conducted literature reviews and benchmarked state-of-the-art CV and DL techniques for aerial imagery.
►Designed hybrid solutions combining traditional CV methods with DL models for accuracy and explainability.
►Developed data preparation and annotation pipelines to streamline dataset creation.
►Implemented and optimized models using Python, C++, and PyTorch for high-performance deployment.
►Adapted and deployed solutions on NVIDIA Jetson devices with real-time constraints.
Mar 2019 — Dec 2019
Paris, France (on-site)
AI Engineer
@ COSE – Aerial imaging systemsDeveloped innovative computer vision and deep learning solutions for aerial image analysis, from research to real-time deployment.
►Conducted literature reviews and benchmarked state-of-the-art CV and DL techniques for aerial imagery.
►Designed hybrid solutions combining traditional CV methods with DL models for accuracy and explainability.
►Developed data preparation and annotation pipelines to streamline dataset creation.
►Implemented and optimized models using Python, C++, and PyTorch for high-performance deployment.
►Adapted and deployed solutions on NVIDIA Jetson devices with real-time constraints.
Robotics Intern
@ ISIR – Sorbonne Robotics LabOct 2016 — Dec 2016, Paris, France (on-site)
Explored robotics fundamentals through the development and control of a robotic arm with sensor integration.
►Calibrated and programmed a Lynxmotion AL5B robotic arm with ultrasonic sensors for object detection.
►Developed algorithms for object retrieval using forward and inverse kinematics with MATLAB and Arduino.
►Contributed to robotics education by demonstrating the system in university courses.
Robotics Intern
@ ISIR – Sorbonne Robotics LabExplored robotics fundamentals through the development and control of a robotic arm with sensor integration.
►Calibrated and programmed a Lynxmotion AL5B robotic arm with ultrasonic sensors for object detection.
►Developed algorithms for object retrieval using forward and inverse kinematics with MATLAB and Arduino.
►Contributed to robotics education by demonstrating the system in university courses.
Oct 2016 — Dec 2016
Paris, France (on-site)
Assistant Engineer
@ EIKEO – Smart Digital displays for advertisingMay 2015 — Aug 2015, Paris, France (on-site)
Gained hands-on engineering experience across product assembly, mechanical design, and workflow automation.
►Assisted in mechanical and electronic assembly of display systems.
►Designed a camera mount using SolidWorks and helped troubleshoot solar simulation panels.
►Automated repetitive tasks using Bash and VBScript to improve workflow efficiency.
►Collaborated with multidisciplinary engineering teams to deliver practical solutions.
May 2015 — Aug 2015
Paris, France (on-site)
Assistant Engineer
@ EIKEO – Smart Digital displays for advertisingGained hands-on engineering experience across product assembly, mechanical design, and workflow automation.
►Assisted in mechanical and electronic assembly of display systems.
►Designed a camera mount using SolidWorks and helped troubleshoot solar simulation panels.
►Automated repetitive tasks using Bash and VBScript to improve workflow efficiency.
►Collaborated with multidisciplinary engineering teams to deliver practical solutions.
Education
Sorbonne University (formerly UPMC)
2019, Paris, France
MSc: Engineering of Intelligent Systems Cursus Master en Ingénierie (CMI): a selective, reinforced university engineering program
►End of Master project: A communication interface for mute people (French to sign language translator)
►Multiple applied AI (ML/DL) projects: image classification, motion imitation detection, voice intent recognition, music genre classification, etc
2019
Paris, France
Sorbonne University (formerly UPMC)
MSc: Engineering of Intelligent Systems Cursus Master en Ingénierie (CMI): a selective, reinforced university engineering program
►End of Master project: A communication interface for mute people (French to sign language translator)
►Multiple applied AI (ML/DL) projects: image classification, motion imitation detection, voice intent recognition, music genre classification, etc
Technical University of Munich (TUM)
2018, Munich, Germany
Erasmus MSc – Robotics & Intelligent Systems
►MSc project in bioinspired robotics: Flexible bipedal walking optimized with a genetic algorithm (Particle Swarm Optimization)
Technical University of Munich (TUM)
Erasmus MSc – Robotics & Intelligent Systems
►MSc project in bioinspired robotics: Flexible bipedal walking optimized with a genetic algorithm (Particle Swarm Optimization)
2018
Munich, Germany
Pierre and Marie Curie University (UPMC)
2017, Paris, France
Bachelor in Electronics, Electrical Energy & Automation
►Bachelor project: Electronic and software design of a secured and automated borrowing system
►Project Romarin²: Full design (electronics, mechanics, software) of a remotely operated underwater vehicle and its robotic gripper
2017
Paris, France
Pierre and Marie Curie University (UPMC)
Bachelor in Electronics, Electrical Energy & Automation
►Bachelor project: Electronic and software design of a secured and automated borrowing system
►Project Romarin²: Full design (electronics, mechanics, software) of a remotely operated underwater vehicle and its robotic gripper
Louis Le Grand Highschool
2014, Paris, France
Scientific Baccalauréat – Engineering Science
►President of the robotics club: Management of equipment and budget, technical support and project supervision
►Engineering Science project: 2D localization and mapping system (SLAM) for apartment mapping
Louis Le Grand Highschool
Scientific Baccalauréat – Engineering Science
►President of the robotics club: Management of equipment and budget, technical support and project supervision
►Engineering Science project: 2D localization and mapping system (SLAM) for apartment mapping
2014
Paris, France
Contact
Decisions come after gathering information.
If you'd like to connect, share an idea, or explore a potential collaboration, feel free to reach out.