1 min read
Videos
The Telemetry Talks YouTube channel, MTTL's educational videos, related projects and the formats planned to extend the articles and labs.
Some things are quicker to grasp when you see them move. A trace, meaning the journey of a request across five services, is easier to follow on screen than in a paragraph. The same goes for a Collector (the program that receives, sorts and forwards observability data) filtering live, or a dashboard where a drift shows up. My videos are published on the Telemetry Talks YouTube channel.
Videos on the channel
Section titled “Videos on the channel”It hosts the first MTTL educational videos, in French or English, along with projects and topics close to MTTL: ReversOtel, Archipel and the security of AI systems.
On this site, each MTTL video sits next to the page it illustrates, in French and English. Observability in 90 seconds opens the home page. From the GPU to the platform sits at the top of the HPC AI course. AI in cross-section goes with its interactive 3D diagram and cardinality explained with the lab that replays the scenario. The HPCOBS lab demo, with English subtitles, sits on the labs page and in the training step of the HPC AI course. Who is growing the metrics bill? follows the monthly report of the costliest metrics and Who builds observability around the world? opens the players map. How do you run the interview? sums up the structured interview and the diagnosis exercise.
GPUs at 87% but how much useful compute? HPC, AI and observabilityYour dashboard shows 87% GPU utilization, yet training is slow.
L'observabilité en 90 secondes : des nilomètres à l'IAQu'est-ce que l'observabilité?
ReversOtel,un controle plane OpenTelemetry open source
ArchipelV0 1
PATTERNS DE DETECTIONAdversarial, Drift et LLM , avec code et metriques.
Threat ModelComment cadrer les menaces sur vos systèmes de ML avant de déployer quoi que ce soit.
Sécurité IA: Pourquoi vos modèles sont-ils de nouvelles surfaces d'attaques?Dans cette vidéo, vous allez apprendre les différents types d'attaque sur un LLM, comment structurer sa protection façe à ces menaces.
Iagentsec/ 1st previewMost companies monitor servers, APIs and users but not the AI itself.
Formats planned for MTTL
Section titled “Formats planned for MTTL”Two formats are planned. The demo, 10 to 20 minutes long, replays a lab step by step. The explainer, 5 to 10 minutes long, explains a concept, an announcement or a standard.
Each video of this kind will link to its article and lab. Every article will link to its video once there is one.