Open to AI infrastructure roles
Network & Systems Engineer Montréal, Canada
Focused on AI infrastructure: GPU clusters, automation and networking, always with security in mind.
About
I'm a network and systems engineer and a student at Polytechnique Montréal. What draws me most is AI infrastructure: the HPC clusters and GPU fleets that modern AI workloads run on. I've designed and operated them with Slurm and Kubernetes, orchestrating NVIDIA GPUs at scale (CUDA, the NVIDIA GPU Operator, and vGPU on vSphere).
That interest grew out of real work: a past role as a scientific collaborator and a run of student projects. It rests on a broader craft too, automation and NetDevOps, cloud computing and a mindset built around security, the foundations that keep large, complex infrastructures reliable. Right now I'm putting all of it into building Glow Up, a platform that brings intelligence to beauty and wellness, and which I co-founded.
Selected work
Designed, planned and implemented the first and only Slurm cluster at HEIA-FR / iCoSys to schedule NVIDIA GPUs (GH200, A40, A6000) used by many data scientists. I configured the full CUDA environment and the NVIDIA Container Toolkit, and mixed bare-metal GPUs with virtual GPUs (vGPU) running in VMs on a vSphere cluster, with cloud bursting from those VMs. The stack is deployed automatically, precisely monitored, and gives users access to three separate storage spaces.
A full rebuild of a GPU-accelerated Kubernetes cluster at HEIA-FR: 3 control-plane nodes and 15 workers, mixing bare-metal machines with VMs on a vSphere cluster, including virtual GPUs (vGPU) and cloud bursting. I deployed and tuned the NVIDIA GPU Operator and configured CUDA with the NVIDIA Container Toolkit to expose NVIDIA GPUs (GH200, A40, A6000), used Ceph as the CSI storage backend, opened access to the school's users through Rancher, and ran several ingress controllers per access policy. It served as the inference platform for models trained on the Slurm cluster I also built.
Part of the European research project FLUIDOS: carbon optimizations through strategic workload scheduling across multiple Kubernetes clusters, leveraging spatial and temporal variation in the carbon intensity of electricity.
The iCoSys applied-research institute runs a large fleet of GPUs. This project analyzed and reported on the efficient use of the cluster, with concrete suggestions for significant improvements. Carried out under a fixed-term contract.
A study of memory management in containerized applications: how the Docker Engine and the Kubernetes orchestrator handle allocation, limitation and monitoring of RAM.
Used Terraform, Ansible and the High Performance Linpack benchmark to measure and compare the compute capacity of various clusters: vSphere, ESXi, OpenStack and Kubernetes.
Bringing intelligence to beauty and wellness
An ecosystem for professionals and clients in the beauty & wellness sectors. For professionals, a free SaaS that supports both the operational and strategic steering of their business. For clients, a mobile app where they discover tailored recommendations and book appointments.
A Tic Tac Toe game written in Go, backed by an SQL API storing game data and a language system. Built for the "Programmation élégante en Go" course at HEIA-FR.
A universe simulator in Java handling multiple types of interacting elements placed on a board, where they move, approach and repel each other. Built for the "Génie Logiciel 1" course at HEIA-FR.
A virtual gym web app: members sign up and carry out exercises, connect with trainers, and pick subscriptions, with member, coach and admin panels. PHP + PostgreSQL on an MVC pattern built from scratch.
Academic path
Polytechnique Montréal
Haute école d'ingénierie et d'architecture de Fribourg
Université Savoie Mont-Blanc
Contact
Open to AI infrastructure, HPC and platform-engineering conversations. The fastest way to reach me is email.
