Paris Open Source Data Infrastructure Meetup - March 2024

Merci à Google de nous accueillir pour cette troisième édition du OSDI Meetup !
Le catering sera sponsorisé par Aiven.
Le programme :
18:00 : Ouverture des portes
18:30-18:45 : Intro d'Aiven et Google
18:45-19:10 : 6 New Ways to Supercharge Your Kafka Experience! (aka a proxy/mesh/sidecar story) , Florent Ramiere, Conduktor
Apache Kafka has revolutionized the way organizations handle real-time data streams. Yet, there's always room for enhancement, don't you want to reduce (drastically) its costs, to make it (way) more secure, and to have (much) less troubles with (clueless) users?
In this session, we'll dive into seven innovative strategies that can take your Kafka deployment to the next level.
Or in other term, we'll examine the reasons that make a kafka proxy an essential companion with tools like kroxylicious, conduktor oss, envoy, greplabs and others!
This talk is tailored for Kafka users who are familiar with the basics and are eager to explore advanced features and integrations.

• Safeguarding: Fix the most common issues, like bad configuration, misusage, and plain abuse.
• Chaos Engineering: Simulate Kafka production failures and check which of your apps fails.
• Encryption: With GDPR, PII etc, encryption and crypto-shredding need to be a commodity.
• SQL Topics: Don't use large processing engine for every small usecase!
• Virtual Clusters: Create many fresh pure virtual clusters with complete resource isolation in your own Kafka.
• Unlimited Partitions: yes, unlimited partitions! (Come see it live to believe it!)

Each section of this talk will include practical examples, best practices, and real-world applications.
Whether you're a Kafka administrator, developer, or data engineer, this session will provide you with a wealth of knowledge and tools to enhance your Kafka experience, making your data streaming processes more secure, efficient, and scalable.
19:10-19:50 : Des modèles open source à la production : intégrer un LLM dans un ensemble d’applications , Amine Saboni - Data Engineer - DiliTrust
Les démo spectaculaires d’IA génératives ont transformé la manière dont les modèles de langages peuvent s’intégrer dans un écosystème applicatif. Entre gestion des performances, de la donnée de fine tuning, et l'exposition d'un modèle multi-tâches dans plusieurs applications, les défis pour s’approprier les LLMs ne manquent pas. Durant ce talk, nous passerons en revue les différentes étapes pour construire sa pipeline de production et de release de LLM, le tout avec des outils open source.