Our Meetups

Past and upcoming meetups.

• AI News #1: Developers, make us vibrate! (2025-03-26) — The first ILAB - AI News: a recurring Meet-Up to discuss AI with you. For all those involved, directly or indirectly, in software development (developers, POs, architects, product managers, UX/UI designers, QA, etc.). Program for this first edition: • AGP - Astek GenAI Platform • Demo of new features, InnerSource, and future perspectives • Key highlights in the world of AI (DeepSeek, Manus) • Vibe Coding and the future of software development professions: presentation, demos, and discussions.
• AI News #2: POA: Agent-Oriented Programming (2025-05-14) — The development world is being transformed by AI in the way applications are built. There's Vibe Coding (see ILAB AI News#1), but there are also agents. Creating AI agents and enabling them to interact with each other is changing our development practices, architectures, and how we interact with applications. Agenda: - AI at Astek - The beginnings of Astek's AI offering - Global AI News - Gemini 2.5 Pro, internal workings of AIs, OpenAI o3, FramePack... - What are AI agents? - Agentic Frameworks (Google ADK, OpenAI Agent SDK, Crew AI...) - Overview, demos, and discussion
• AI News #3: From the MCP Revolution to Agent 2 Agent (2025-07-02) — The Model Context Protocol (MCP) is already shaking up how AI agents use services or applications, delegate tasks, and integrate into our architectures. Following and alongside Vibe Coding, get ready to discover the era of interconnected agents! 🔥 Agenda: Flash Global AI News — Overview of the AI world Claude 4 and beyond Vibe Coding Kilo Code - What is the MCP protocol? - Principles and Specifications - When to use MCP? - Key MCP servers to know - What is Agent 2 Agent? - Principles and Specifications - MCP / A2A - Live Demo - Connecting to an MCP server - MCP + AI Agent + Inter-agent Communication - Debate & Q/A – Your questions, our practical insights
• AI News #4: From RAG to Graph RAG (2025-09-24) — From RAG to Graph RAG—the evolution of Retrieval Augmented Generation. By structuring knowledge as graphs, this approach brings context, connections, and relevance to your AI systems. Come discover how this new paradigm is transforming the way we enrich and leverage data. 🔥 Agenda: - Flash Global AI News - Overview of the AI world - First Encounter with Context Engineering - Introduction to Context Engineering - A training session? - An Awesome resource? - From RAG to Graph RAG - Differences between RAG and Graph RAG - Context & Methodology - Architecture - Live Demo - Graph RAG vs. RAG in action - Discussion & Q&A - Your questions, our practical insights
• AI News #5: AI IN REAL LIFE (2025-11-19) — Today, AI is making its way into every aspect of our professions. During this Meet-Up, we will explore how AI helps us support our clients, with a concrete use case developed by the Rennes team. We will also examine how AI is transforming the role of UX/UI Designers, thanks to Valérie Kowalski, who didn't wait and seized the opportunity to master AI, boosting her productivity and showcasing one possible path for the evolution of her profession. 🔥 Agenda: - Flash Global AI News - Overview of the AI landscape - RAG in real life - Project context - A bit of architecture - Live Demo - The new UX/UI Stack - From classic tools to AI tools - Magic Path Discovery - Live Demo - The New ILAB PoC - Jerôme Gisson presents MCP for devs PoC - Debate & Q/A - Your questions, our practical insights
• Pulse Meetup #1: From ARIMA to Transformer - The Forecasting Revolution in 2026 (2026-04-30) — > How generative AI and Foundation Models are reinventing time series forecasting — supported by live demos. ### The core of the meetup For decades, predicting the future of a time series (sales, energy consumption, engine wear...) relied on proven statistical models like **ARIMA** or **Prophet**. Effective, yet limited when it comes to combining multiple variables, explaining a prediction, or adapting to a new context without retraining from scratch. In 2025–2026, forecasting is undergoing its own **Transformer revolution**. Architectures like the **Temporal Fusion Transformer (TFT)** bring attention mechanisms and interpretability to the heart of forecasting. And above all, **Foundation Models** (Amazon's Chronos, Google's TimesFM) now enable predicting a time series **without any prior training** — just like an LLM can answer a prompt it has never seen before. This meetup is a **hands-on, accessible dive** into these new approaches. No intimidating math: just code, graphs, honest comparisons between the old and new worlds, and industrial use cases (predictive maintenance, demand forecasting). ### Target audience Developers, data scientists, architects, tech managers — and anyone curious to see the state of AI applied to forecasting in 2026. **No advanced time series expertise required**: we start from the basics. ### What you'll take away - A clear vision of the 2026 forecasting landscape (and what's truly production-ready) - Reusable notebooks to kickstart your own experiments - The "zero-shot forecasting" mindset for your next POCs
• Pulse Meetup #2: From AI That Talks to AI That Acts — MCP Apps & Self-Improvement Harnesses (2026-06-04) — > When the model delivers the interface and the application improves itself — backed by live demos. ### The Heart of the Meetup For two years, we've learned to **talk** to AI. Pulse #1 showed how a Foundation Model can predict a time series without training. Six weeks later, we're taking the next step: making AI **act** within our applications. Two trends are converging in 2026. On one side, **MCP** (Model Context Protocol)—the "USB-C of AI," opened to the entire industry by the Linux Foundation—is becoming the standard for giving models access to tools, data, and now **real interfaces**. With SEP-1865, an MCP server can return an interactive UI (form, dashboard, visualization) rendered directly in the client. We call these **MCP Apps**. On the other, **self-improvement harnesses** applied to applications. Instead of asking an LLM to write code with full autonomy (a risky utopia), we place it within a **disciplined loop**—measure → propose → validate → iterate—around an existing app. Targets include performance (profiling, optimization), security (scanning CVEs, patching, retesting—generalizing the Claude Mythos / Glasswing pattern), and quality & UX (testing, refactoring, continuous A/B). **LLMs are the tool, not the target.** This meetup is a **hands-on dive** into these two new layers of tooling, with a strong focus on self-improvement harnesses. No hype, just code, live demos, and an improvement loop running before your eyes. ### Target Audience Developers, data scientists, architects, tech managers—anyone maintaining an application and wondering what AI agents can concretely bring to it. **No prerequisites** on MCP or agents: we start from the basics. ### What You’ll Take Away - A clear vision of the MCP Apps landscape in 2026, and how it can serve you (and what’s truly production-ready) - A mental model for deploying a self-improvement harness around an existing app (and identifying essential safeguards) - A live demo of an MCP Server + UI - A live demo of a Self-Improvement Harness