Suggestions
Pierrick Pochelu
Deep Learning Data Scientist & HPC Data Engineer @ Total / CNRS
Professional Background
Pierrick Pochelu is a highly skilled and knowledgeable expert in the fields of Machine Learning, AutoML, and High Performance Computing. His journey through academia and industry illustrates a commitment to research and technological innovation. Since embarking on his career, Pierrick has cultivated a wealth of experience in Deep Learning research and development, which has positioned him as a key player in advancing computational methodologies across various applications.
One of Pierrick’s notable contributions involves the automation of microscope counting utilizing deep learning techniques. He began this project in 2016 and succeeded in developing multi-GPU training tools and a fast inference engine, showcasing his ability to tackle complex technical challenges head-on. Additionally, Pierrick has made significant strides in biodiversity research by implementing automatic monitoring systems for jungle animals using camera-traps, deploying a weakly-supervised Faster-RCNN method starting in 2019.
With an eye for innovation, Pierrick is currently focused on designing and implementing a groundbreaking multi-objective AutoML software tool. This sophisticated software, which operates under specific constraints such as hyperparameter optimization (HPO) and ensemble machine learning, has been successfully validated on diverse tasks within both Computer Vision and Reinforcement Learning. This work began in 2020 and underlines his commitment to enhancing machine learning frameworks using multi-node GPU high-performance computing (HPC).
Moreover, Pierrick has ventured into the realm of Deep Learning optimization. In 2018, he benchmarked gradient-free optimizations, providing valuable insights into training efficiency and model performance for various applications. Beyond his technical expertise, Pierrick has contributed to the academic community as a reviewer for the International Conference on Computational Science (ISSC), where he ensures the integrity and quality of applied machine learning research.
Education and Achievements
Pierrick Pochelu has an impressive educational background that laid the foundation for his career in technology and computation. He pursued a Master recherche in Computer Science (Mention B) at Université de Pau et des Pays de l'Adour, where he honed his analytical and technical skills. Additionally, Pierrick earned a Diplôme d'ingénieur in Computer Science from the Ecole internationale des Sciences du Traitement de l'Information, equipping him with a robust understanding of data processing and computational sciences.
His educational journey is complemented by practical experience gained through various roles where he engaged in research and software development. As a PhD student and software engineer under a CIFRE French contract at Total & CNRS, Pierrick was able to connect academic research with industry needs, driving forward innovative projects.
Pierrick's earlier roles as a Data Scientist in Deep Learning at both LumenAI and Total provided him vital hands-on experience in real-world applications of machine learning. His tenure as an R&D intern at Total, where he focused on 3D visualization and heavy signal compression algorithms, further developed his technical proficiencies. Pierrick also interned as a web developer and software engineer, expanding his skillset across diverse IT domains.
Achievements
Over the years, Pierrick Pochelu has solidified his reputation as a forward-thinking expert through various successful projects and innovations. His contributions to the fields of machine learning and data science are marked by a few key achievements:
- Automation of Microscope Counting: Pioneered a project utilizing deep learning for automating the counting process with the development of multi-GPU training tools, enhancing efficiency and accuracy.
- Biodiversity Monitoring: Implemented automatic monitoring systems using innovative camera trap technology to study jungle animal populations. This project not only showcased his technical skill but also highlighted his commitment to ecological research and conservation.
- Multi-objective AutoML Software: Developed and tested an AutoML software capable of conducting hyperparameter optimization under various constraints, marking a significant advancement in the machine learning landscape since its introduction in 2020.
- Benchmarking Gradient-free Optimizations: Contributed to the optimization of deep learning processes by benchmarking various approaches, providing insights that have informed future developments in the field.
- Reviewer for International Conference on Computational Science: As an established reviewer, Pierrick has actively contributed to the vetting and selection of research papers, reflecting his expertise and professional integrity within the academic community.
Beyond his technical innovations and research achievements, Pierrick has embraced the role of educator as well. He has trained students at EISTI engineering school, delivering a comprehensive 42-hour course on GPU scientific programming. This initiative emphasizes his dedication to sharing knowledge and fostering the next generation of tech professionals.
Overall, Pierrick Pochelu stands out as a dedicated and innovative professional in the fields of machine learning and high-performance computing. His blend of technical skills, research experience, and educational contributions make him a valuable asset to any organization or project in the tech landscape.