If you think you have it tough, look around.
Enthusiasm for
Minimalism (only using Debian, i3wm, Bash, vim with plugins), autonomy (draft my notes on memorandum and memos, build my own daily scripts with mylinux†, store my projects on Gitlab, test codes that I store on my sandbox, manage my tasks with Vikunja, follow my feeds with miniflux), automation (Ansible/Salt, ML with K8S, Scalable Slurm Cluster for Cloud), Python, C, optimization (complexity, meta-heuristics), security (manage my own self ca stack, isolate the apps: Docker/Podman, privileged Flatpak for GUI, libvirt for services; With a degree in the field, I keep an eye on misc), research in AI (classical ML, DL*, interpretability, see my thesis manuscript, my research articles: tpot-sh, autotsc, or my collection of articles), Rust*, OpenBSD*, Quantum Algorithms*
†bunch of tools that ease daily usage of CLi that I plan to opensource
*mid/future plan, reading about it, but not necessarily practicing
Self host
I like to self-host a bunch of services, which comfort me in where my data are stored and how it is manipulated. It also allow me to freely configure the services according to my need (resources, plugins, reverse proxy). See below to have an overview of my personal stack (might be outdated):
Personal network
I also like to handle my own network (router, switch, AP) for different reasons (security concern, keep practicing my basics in networks). Thus, I choose different devices that are compatibles with OpenWRT (an open source linux based distribution oriented for network devices), see discussion for more details. In a very short summary, I have a Ubiquiti edgerouter-X, 2 * Netgear GS108T, and 3 nodes serving as APs (mesh because I can't have holes in the current house) that are Unifi LR-6/Lite6. Everything is managed through Ansible or CLI such as I can easily scale my network by adding new nodes or new rules: firewall, VLan, SSID, bridge. For now, I've isolated the network in 5 parts: lan, service, IoT, home, gaming, guest. I have also added a failover link such as if the main ISP is down, IoT and Lan will still operate on a second ISP.
Ideally I would like that each device is based on a low powered open hardware solution (ARM board, e.g. RockPro64), with OpenBSD. However it requires quite some energy/time/maintenance (even more costly than mainstream solutions) to have such a setup i.e. build OpenBSD for each ARM board (depending if its a router, switch or AP), configuring it such as I can operate the current setup as well, i.e. VLan, SSID with Wifi6, etc. Maybe I will do that during my retirement, or when my kids will grow and will have some interest on their daddy hobbies, if they do.. who knows 😄
Aside I have multiple ARM boards, such as Rock64 for which I've built a PicoCluster and a Rockpi4B. It was initially to self host my services, but with time I realized there were to many constraint to maintain such a setup at home (in few words: unexpected power outages, network disruptions from ISP, slow I/O with SD cards, OS issues during upgrades: ARM community is smaller than x86). I still keep them for entertainment, and I still believe that they remain interesting in a near future (e.g. low power, new boards have PCI-e: improve I/O by replacing SD card with NVMe drives)
Current position - Data Scientist/MLOps
Currently at OVHCloud, I was a DevOps first to develop an AutoML that to solves internal problems (e.g. fraud detection, HDD anomaly detection). However I wasn't so interested by the DevOps part, but more by what is under the hood of an AutoML (core optimization). Luckily for me, there was an opportunity to undertake a CIFRE Thesis†in that field. During that opportunity, I also switch my position to Data Scientist (out of contract, its more like MLOps), which fits way better with my aspirations. Roughly speaking, I develop and maintain machine learning models for different internal use cases (Forecasting Time Series*: 600k TS, DDoS L7 mitigation: already 328 attacks in 2023 ranging from 10 to 90k req/s, Product Recommendation API: 1.6M customers, Waterblock Anomaly Detection: 400k servers monitored, hundreds of waterblock changed/year which saved e.g. 800k€ in 2022). I also developped some new concepts through patents: autosea, dsop, [AA] (ongoing)
†specific contract of thesis which is done between a public laboratory and a private company
*ongoing
Past experiences
Fullstack Dev / Webmaster
Concerning very first experiences, let say from 12 to 20 y.o., I've had the chance to build many websites (french singer, ski club assocation, firm of funeral directors, association helping poor elderly persons, network helping foreigners, audiovisual production company) by using different technologies - depending to the need (CakePHP, Symfony, CodeIgniter, Home made solution, and Wordpress), yes, the famous LAMP world. Please not that I do not appreciate to work with these technologies anymore, dont try to hire me for that, I'm bored. Nonetheless, it was a very first great experience to deal with PHP community and its emerging concepts that I've discovered (MVC pattern, ORM, templating).
Sys. Admin
During my free time, I've also maintained the infrastructure of winescholarguild.org (frenchwinesociety.org once) for years (from my 14 to 30 y.o.). I've seen the website growing from few hundreds users to more than 30k+. The website is composed of two frameworks: Joomla (e-commerce) associated with many plugins and a Moodle (courses, exams). I've ensured that the infrastructure meets all their need. I've automated the installation and the configuration of the stack with Salt (Apache for HTTP, Bind9 for DNS, Proftpd for FTPS, Mysql for the database, Fail2ban for the firewall, etc). I've also ensured their Business Continuity Plan, by managing NAS with daily snapshots and backups concerning the databases in addition to website contents. In addition, the company tends to send hundred thousands of emails per month (newsletters), to avoid being in spam, I had to install and configure a dedicated mail server from scratch (Postfix, ClamAV, Rspamd, DKIM, etc). Thanks to containers world, we recently migrated with a better solution in terms of maintenance, security, and functionality: Mailcow.
Nowadays, I'm in a full time employment and I have no more time to maintain it, however I'm still like a consultant by advising their choose of technologies (framework, hardware requirements, update processes, emergency failures), I'm deeply thankful that they did trust me to manage all their infrastructure, it helped me to practice and touch a lot of stuffs.
DevOps
One of my most notable project concerns the complete set-up of a High Availability (HA) and Scalable cluster that sync files between computers (Nextcloud based application). We did the whole setup in less than 3 weeks with my very close friend Jonathan Franck, that taught me so much aspects of Sys. admin and DevOps (I owe you so much dude). In a summary, we developed all the rules with Salt such as we can add and configure every new nodes at every level of the stack. We managed our own advanced DNS records (Glue Records, DNS Robin Round). We used KeepAlived for HAProxy based load balancers, such that if a loadbalancer is in failure, the second one retrieve all the TCP connections. We put the load balancers on top the of frontend nodes (HTTP requests), and on top of the backend nodes (for MariaDB master nodes, and Memcached), such that all the network flow is well distributed. We also used GlusterFS to scale our storage facility. Technically, it was a very solid project, however it stopped due to legal issues. Aside I also developed a client and a server backend in order to run tasks asynchonously with external libraries.
Quality Assurance Engineer
I had an awesome opportunity to work for Intacct, in the heart of the Silicon Valley (San Jose). One of my greatest experiment in life (visiting California, meeting passionate techies: Ledger's CTO, Docker's CTO, researcher at Criteo, and so more thanks to while42 network). Honestly, when I came, I barely spoke english.. Luckily for me, I've been fastly integrated by very kind persons for those I'm still grateful today !
During this experience, I was part of the Quality Assurance and Release engineer team. It was great to learn how we ensure the continuity of live releases, such that we minimize the bugs in production (unit tests, integrations tests, API tests). I used this period to learn more about CI/CD concepts, more specifically with Jenkins and Gitlab-ci (e.g. run dedicated container environments for each branch). Concepts that still serve me for my current tasks, e.g. put in prod with CDS at OVHCloud, or to build and deploy some of my webpages such as memorandum through my own instance of Gitlab. During that job, I also improve the internal home made tools used by QA Engineers to increase their productivity when they compare the results of API calls between two releases. I don't want to go too much in details, but I spend my time on other aspects that are unrelated to QA and that I enumerate on my LinkedIn page or in my resume (e.g. learning about AI: ML certificate (classif, regression, clustering), DL certificate, build a POC TS anomaly detection based on S-H-ESD).
Hobbies
Boxing (weekly practicing), Stunt (long time ago but I still have some interests, please note that it was done in private roads with permissions, I dont support those who put general public in danger), Snowboard, Motocross (I had a YCF Pilot 150F, I had to stop since I live in town), Hiking, Old School Games (Counter Strike, Killing Floor, WOW ofc like many, but I'm too old now, it takes a life to play), Trips/Travels (USA/Canada), Tuning things/DIY (e.g. scooter engine, watercooling computer), being crazy with my two kids like dancing Daddy Cool 😄, and listening songs (Yoke Lore, Roo Panes, Of Monsters and Men, Hey Nothing, The Kiffness)
Mad skills
Despite a very tough history (repeat a year at school, refused on scientific program at High School, only student to retake the final exam at High School), I fought hard along years to finally obtain two Master degrees. I went out from my comfort zone multiple times: expatriation in Canada (Sherbrooke) and USA (San Jose). Both time was for a job, not for vacation. I did different kind of programs: J-1 Visa two consecutive times, EIC: Jeune Pro + Stage COOP). I did a PhD, which was a first for OVH, considered as the biggest cloud provider in Europe, in a field that I knew very few (Machine Learning), and with my first kid born at the very beginning of the thesis - 4 intense challenging years, big thanks to my company advisor who actively participated in the success of this challenge. I'm subject to cluster headache each 2 weeks (family issue), and still, I'm dealing with tough challenges. I hope its mad enough to show how I bounce and I can adapt myself to situations !