Available for a final-year internship starting February 2027

Hi ! I'm Mathis Delmaere, a computer engineering student at the Université de Technologie de Compiègne (UTC), in the Information Systems Engineering (ISI) track (Class of 2022).

I code mostly to learn, explore new technologies and computer systems (backend, infra, or networking), or to build things for student associations at my school.

• A quick intro?
• Professional experience
• Flagship project - k3s Homelab
• Student association life at UTC
• Coursework projects
• Hackathons
• Stack & Technologies
• Contact

At barely 10 years old, I was already spending my time playing Minecraft, digging through the infamous %appdata% folder to customize my game, installing modpacks that were absolutely not meant to work together, and tinkering with Java and VMs to try to understand how the game worked and how it ran on a PC.

The logical next step was to build my own PC to run a souped-up Minecraft, install Linux on it because it's cool, start doing development, and then decide I'd do this for the rest of my life.

Throughout my time in student associations and academics, I've genuinely had a blast with web/mobile development, but I'll admit I don't enjoy it as much as I used to. Ever since I discovered DevOps, infrastructure, design patterns for scalable and resilient backends, and more broadly everything related to SRE, I feel like I've found a much bigger and more interesting playground than plain development.

Infra & DevOps work at a French occupational health scale-up.

• Kubernetes / ArgoCD: maintaining and evolving multi-environment Kubernetes clusters (staging/prod), managing GitOps deployments
• Monitoring: setting up and operating Grafana dashboards, Prometheus queries, long-term data aggregation with Thanos
• CI/CD: writing and maintaining GitHub Actions pipelines
• Go: developing internal infra tooling

• Academic support tutor in mathematics and computer science · UTC, Compiègne: September 2023 → June 2025
• Production line operator intern · Chanel, Compiègne: February 2023 → March 2023
• Freelance website development: January 2023 → March 2023

A k3s cluster I've been administering end-to-end since November 2025: this is the playground where I put everything related to infra, networking, and SRE into practice, and where I run several of the association projects presented below in production.

Full repo (GitOps, feel free to clone and explore): mathisdlmr/k3s-project

The HA k3s cluster runs across 3 little computers (NUCs) spread across 2 different homes, interconnected using a Tailscale (VPN) mesh. Cilium routes its inter-pod VXLAN network through this Tailscale tunnel, and etcd handles distributed consensus with automatic snapshots every 6 hours. Longhorn is also set up to replicate volumes across every node, so storage stays accessible even if a node goes down.

PS: The 3 NUCs were originally all at my parents' place, hence the photo above, but I later moved 2 of them to my apartment in Compiègne

On my machine, a local HAProxy round-robins across the 3 control planes for HA access to the API server: if one node goes down, kubectl, ArgoCD, and all my other services keep running without interruption.

For the cluster, I went with a DevOps approach that's fully reproducible in just a few minutes.

The cluster is first bootstrapped by Ansible, which runs on my laptop and configures each NUC to become a (secured) node

Once the setup is complete, this is the resulting architecture:

The next step is installing an ArgoCD application that takes over management of the ArgoCD instance installed by Ansible, and handles deploying the rest of the repo through the meta application

Finally, all that's left is to pull in the secrets. For that, I create a Kubernetes secret called infisical-universal-auth-credentials, which lets External Secrets Operator authenticate with Infisical and then recreate every secret defined in the repo (logins, Cloudflare tunnel, etc.)

And there it is, the cluster is rebuilt from scratch!

The services are then reachable online at https://mdlmr.fr/*. And for a bit more visibility into how that works, here's one last diagram ;)

The cluster is driven by ArgoCD following a multi-level app-of-apps pattern (with sync-waves to guarantee deployment order: ArgoCD itself and its CRDs first, then infra, then monitoring, then the apps), and by Renovate, which automatically opens PRs to update Helm charts and Docker images (auto-merged for minor updates, manually reviewed for major ones).

A GitHub Actions workflow, Argo Diff Preview, generates the full manifest diff (rendered Helm + Kustomize) and posts it as a comment on every PR. Handy for seeing the impact before merging :)

Observability is handled by 2 complete stacks:

• The Prometheus and Grafana Labs stack (Prometheus, Grafana, Alloy, Loki, Tempo), as well as
• The EFK stack (Elasticsearch, Fluent, Kibana) - Victoria Metrics - OTel

Both monitoring stacks collect logs and metrics from all deployed services, while the tracing system is currently only used for the Ski'UT backend, where I implemented the OTel SDK to collect traces and inspect particularly slow requests (the bottleneck was usually database connection pooling or poorly factored Eloquent ORM operations).

The point of running 2 monitoring stacks is purely for learning purposes, so I'm comfortable with the Grafana Labs stack as well as other widely used tools (Elasticsearch, Fluent, etc.). Along those lines, I'm also planning to add Mimir and/or Thanos IO to my stack soon, along with Datadog.

In January 2026, this cluster handled load spikes in production from the Ski'UT booking mini-game - averaging around 150 req/s and peaking at 200 req/s - absorbed thanks to a Cloudflare cache configured on storage (mainly images, CSS, and JS), Traefik running as a DaemonSet in front of the cluster, and load balancing tuned by an HPA that could scale from 1 to 3 backend containers under heavy load.

Under normal circumstances, the server also hosts my personal services (Affine as a Notion-like tool, Immich for photos, my personal website, etc.).

Took over, debugged, and updated (new features, new design) a Flutter mobile app and a Laravel backend for a 1.5-week orientation week. The app was used by around a hundred students and was used to run activities and manage the organization of orientation week.

Took over, debugged, and updated an Expo mobile app and a Laravel backend for a 2-week orientation program. The app was used by more than 1,000 students and had even more features than the Integ Fev app: meal booking, virtual queueing, shuttle booking, QR code scanning, and more.

These screenshots are from the version of the app I started working on, where I fixed bugs and added features. The initial development was done by Géo SAGLIO the year before I joined the association. Unfortunately, I wasn't able to track down older versions of the project, since the repository is private and I don't have a local copy...

Built from scratch (with my roommate at the time, Eric BJARSTAL) an Expo mobile app and a Laravel backend to organize a ski trip for ~500 students and run activities throughout the week.

• Backend: ski-utc/server-skiut-2026 - Laravel/Filament server handling the entire trip organization (auth via the student union's IT department OAuth 2.0, bookings, schedule, shuttles, etc.)
• Mobile app: ski-utc/app-skiut-2026 - Expo app with challenges, schedule, fun facts, resort map, shuttles, push notifications, GDPR export/anonymization, etc.

View the project presentation

Took the project back over, this time solo, to stabilize it and make it sustainable long-term:

• CI/CD for both frontend and backend (dockerization + backend unit test pipeline, ESLint + Prettier for the frontend)
• Deployed the backend in self-hosted Docker containers on my personal Kubernetes cluster (→ see the k3s Homelab section above)
• Integrated metrics and traces into the backend, collected by the cluster's Alloy agent
• Wrote documentation explaining how the project works and the best practices to follow

The project has since been running entirely on my Kubernetes cluster.

UTC's student bar and common room. I worked on maintaining the IT systems and developing new features for the treasury and events teams (Spring 2025):

• Ocktopus - picasso-utc/ocktopus: Laravel/Filament backend for organizing the association and treasury services, plus an API for a future mobile app
• Bach - picasso-utc/bach: a React-based payment kiosk installed on Raspberry Pis connected to an NFC badge reader, which reads student card UIDs through a Java applet running on the Pi, so the React app can finally process the sale through the "Weezpay" service
• Overhauled the documentation, updated the projects, and migrated the sales kiosks and display screen from Raspberry Pi 3 to Raspberry Pi 5

IT Department of the Student Union - hosting and infrastructure for the 100+ associations in the BDE-UTC federation.

• UTCats: Filament webapp for managing CATs - mathisdlmr/UTCats
• Debugging and development on infrastructure projects (mostly private)

Elected president of Le Pic'Asso, UTC's student bar and common room, for a 6-month term leading a team of 25 people.

• Main point of contact with the student union's presidency, the school administration, the health and safety department (SHS), and the fire department
• Managed facility-related projects: planning, grant applications, construction follow-up
• Worked on CSR issues, community living standards, non-discrimination, and combating sexual and gender-based violence
• Provided individual support to association members to ensure their well-being and the smooth running of the common room
• Organized near-weekly events, each with its own safety file
• Oversaw logistics (~1,000L of beer, 125 sandwiches and 250 pastries per week, 400 cans of soft drinks) and the related supplier relationships
• Maintained the premises and IT systems, supported the association's communications, etc.

Beyond computer science courses, I also took a number of courses in cognitive science, as well as on the relationship between technology and cognition. In terms of qualifications, this corresponds to earning the PHITECO minor (PHIlosophy, TEchnology and COgnition) alongside my degree.

"PHITECO offers scientific, philosophical, and practical tools for understanding how technology transforms the way we think, perceive, reason, act, and interact. The minor introduces engineering students to the major theoretical and practical questions of cognitive science."

To give a simple example of what PHITECO covers, you'll find here a paper I wrote with my classmate Lysandre FINTA--LAURENT for the PHITECO minor seminar. The paper examines how psychiatric classification systems (DSM-V, ICD, RDoC, etc.) affect their own use across different application fields (insurance, education, research, etc.). The paper's throughline draws on Jack GOODY's work on graphic reason and Bruno BACHIMONT's work on computational reason: writing and computing are not neutral tools — they shape the range of what's possible in terms of action and cognition.

A mobile app that gamifies cultural discovery: a map of cultural venues (via OpenStreetMap), quests, podcasts (via PodcastIndex), books (via the Google Books API), and a shop to spend earned XP. Built in 48 hours.

Stack: Expo · React Native · TypeScript → mathisdlmr/CultureXP

View the project presentation

A local peer-to-peer support platform for students: item lending, service exchange, knowledge sharing. Designed with accessibility and eco-design in mind (Server Components, optimized Prisma queries, static rendering, Positive Impact Score).

Stack: Next.js · TypeScript · Prisma · TailwindCSS · NextAuth.js → mathisdlmr/hackhaton-utc-mc2i

View the project presentation

Coming soon:

• Associations: Pic A26? TX Kubernetes?
• Projects: Cyber-resilience in SR07? Cloud in SR08?

• Email: mathis.dlmr@gmail.com
• LinkedIn: linkedin.com/in/mathis-delmaere-6a6325325
• GitHub: github.com/mathisdlmr
• Location / mobility: open to relocating internationally or working in France with international teams
• Languages: French (native), English (C1)

Thanks for stopping by!