DevFest Toulouse 2025
2025-11-13
Labège, France
[DevFest Toulouse 2017, DevFest Toulouse 2018, DevFest Toulouse 2019, DevFest Toulouse 2023, DevFest Toulouse 2024, DevFest Toulouse 2025]
as the organisers put it
Retour sur le DevFest Toulouse 2025 : conférences, speakers, chiffres de fréquentation, aftermovie et temps forts de cette édition.
| title | author | length |
|---|---|---|
| Aftermovie | 03m | |
A conference aftermovie compiles interview clips and b-roll that focus on atmosphere and networking rather than technical detail. Footage highlights One Point… | ||
| Short | 01m | |
Seven spoken instances of the word "Heat" punctuate a minute‑length audio piece layered over electronic backing marked as [music], with no technical claims… | ||
| 45 minutes to bring your app to its knees the complete guide to load testing | Loïc Ortola, Mathilde Lorrain | 44m |
Builds a capacity test for the Taxini web platform focusing on the public gateway using Gatling (run locally with the Maven Gatling plugin and with richer… | ||
| Under the Hood of LLMs — all those questions you never dared ask about them | Guillaume Laforge | 46m |
Describes why large language models are trained on enormous corpora (about 13 trillion tokens) and why they use subword tokenization rather than characters or… | ||
| One Pixel Then Another: How to Run DOOM on an E‑Reader | Moustapha Agack | 46m |
A developer shows how they ported the Doom engine to a jailbroken Kindle e‑reader: they used a purchased WAD and compiled Doom Generic for an ARM Cortex‑A8… | ||
| Codepocalypse Now Lang Chain4j vs Spring Al | Baruch Sadogursky, Viktor Gamov | 46m |
Both frameworks implement the same airline-loyalty assistant (OpenAI APIs, PDF knowledge base, embeddings) and demonstrate LLM calls, conversational memory… | ||
| Kafka 4, Fantastic? | Damien Lucas | 46m |
Kafka 4 introduces around 40 KIPs (18 in 4.1) that require real migrations: clients and Streams now need Java 11, brokers and Connect need Java 17, and old… | ||
| From Zero to Data as a Product in 60 Days: Anatomy of a Successful Data Product Launch | Maria-Eliza Paez | 42m |
A Michelin Simix team of four describes launching a paid data-as-product from zero to first sales in 60 days by running a 3.5-day design sprint, validating… | ||
| Unbridled Home Automation, from Trash Bag to Wild Boar | Sylvain Wallez, Erwan Rougeux | 49m |
Describes a DIY home-automation setup built around Home Assistant on a Raspberry Pi with Zigbee and many opt‑ins to run concrete automations. Examples include… | ||
| At the Heart of a Core Team | Alexandre Daubois | 41m |
Symfony core team manages a large repository (32 billion downloads, ~4,000 contributors, ~300 open PRs) using a custom GitHub merge CLI that requires two core… | ||
| Docker Bake, elegance and standardization for building your Docker images | Mazlum Tosun | 45m |
Docker Bake (GA 2025) standardizes Docker image builds by using BuildKit/Buildx and an HCL-based configuration to replace fragile shell scripts, offering… | ||
| Untangling Real Products and Hallucinations, Postmortem of a Phone Agent Sock Finder | Marie Terrier | 15m |
A phone-agent for finding items in a 20,000‑product medical hosiery catalog faced ASR errors on technical terms and LLM hallucinations once candidate lists… | ||
| Beyond agile Behind the scenes of the Tech giant unicorns | Rachel Dubois | 46m |
Describes how large tech companies organize delivery with a Spotify-style stack: Backstage and Tableau for dashboards, Figma for design components, feature… | ||
| Web Haute Couture: The Great CSS Catwalk | Audrey Gentili, Etienne Idoux | 47m |
Compares several ways to build web styles: plain CSS (selectors and files) that scales poorly; Sass (variables, nesting, mixins, modules, 20‑year ecosystem)… | ||
| AI and blockchain, allies of security | Carmen Piciorus | 14m |
Demonstrates how AI (voice cloning from as little as 3 seconds and attacker models like WormGPT) and blockchain technologies can be used for security: shows a… | ||
| Postmortem: A Modern Approach to Managing Kubernetes Clusters | Alexandre Dias, Thomas Martin | 47m |
Conto replaced Cluster Autoscaler with Carpenter to reduce node fragmentation and reclaim capacity, runs Argo CD GitOps using an app-of-apps bootstrap to… | ||
| Tech Watch: My Method for Staying Up to Date | Camille Roux | 11m |
The talk explains a workflow for replacing social media feeds with RSS readers to regain control and stay up-to-date, recommending Inoreader (with mentions of… | ||
| Demystifying OAuth2 and OIDC | Nejmeddine Ben Ouarred | 45m |
Explains OAuth 2.0 and OpenID Connect as standards-based authorization flows, naming endpoints (/token, /introspect) and roles (resource owner, resource… | ||
| Another World, a Beautiful Lesson in Software Architecture | Olivier Poncet | 45m |
Rebuild of the 1991 Amiga game Another World from its original executable and assets, describing the tiny 24 KB binary and 1.2 MB assets on a single floppy and… | ||
| Weak signals of a vocal revolution in UIs | Simon Gomez | 16m |
The talk describes how voice interfaces are shifting from novelty to practical UI components, driven by large conversational models and platform integrations… | ||
| Discovering the BERT family of models | Alexia Audevart | 43m |
An overview of the BERT family, explaining that BERT uses Transformer encoders to create contextualized embeddings and is trained with Masked Language Modeling… | ||
| MCP and A2A Revolutionizing Interconnection and Collaboration of AI Agents | Laurent Grangeau, Tony Jarriault | 46m |
MCP defines a standard for LLM tool calls (STDIO or streamable HTTP), adds OAuth and an on‑behalf‑of flow for per‑user backend access, and exposes tools… | ||
| Interactive Learning and User‑Centered AI: A Retrospective | Kevin Delcourt, Trouilhet Sylvie | 13m |
Presents OCE 2024, an opportunistic composition engine that discovers local service components and assembles them using a multi-agent choreography trained with… | ||
| When AI Meets High Performance Computing | Alejandro Estana | 46m |
Practical guidance on running deep learning workloads on HPC clusters using PyTorch and Jupyter frontends with Slurm job scheduling and container runtimes like… | ||
| Draw me a software or the art of diagrams that stick in the mind | Fabien Marty | 15m |
Argues that diagrams should favor clarity over completeness by removing clutter and explicitly noting what was omitted; shows real before/after examples where… | ||
| Behind the Scenes of JavaScript: What We Use Without Understanding | Etienne Idoux, Mickaël Alves | 44m |
Covers JavaScript implementation details from a 1995-to-present timeline, naming engines V8, SpiderMonkey and JavaScriptCore and explaining how engine, runtime… | ||
| The Invisible Monster: The Carbon Footprint of AI | Olivier Bierlaire | 44m |
Generative AI's training and deployment already consume large amounts of electricity and CO2, with a cited rough estimate of about 1 tonne CO2 per billion… | ||
| Renovate to the Rescue | Paul Vulliemin | 16m |
Renovate is presented as an automation tool that scans infrastructure-as-code and CI files—Argo CD Helm apps, Helm values, Docker image tags… | ||
| Metal As A Service — Manage Your Bare Metal in MaaS Like a Virtual Machine | Julien Briault | 41m |
Shows how Canonical MAAS is used to run a self-hosted mini-cloud that treats physical servers like VMs across three regions (Chartres, Paris, Marseille)… | ||
| Critical UIs: When User Tasks Test the Code | Celia Martinie | 27m |
Presents a method where hierarchical task models drive automated tests of a live UI by co‑executing alongside application code using a modelling IDE with task… | ||
| Orchestrating a Role-Playing Game Universe with Multi-Container AI Agents | Guillaume Lours, Philippe Charrière | 01h 40m |
The talk describes building a containerized multi-agent RPG using small local LLMs (0.5B–15B) fine-tuned on ~6–12k QA pairs, combined with RAG and… | ||
| The Hell of Automated Tests | Jonathan Gaffiot, Julien Lenormand | 44m |
Practical advice for building reliable, fast automated test suites used at companies like Google and Netflix: aim for 5–15 minute feedback by splitting suites… | ||
| Talk to your data without coding agentic AI and Claude with the MCP protocol | Han Héloir | 28m |
MCP is a small server that exposes database tools so LLMs like Claude or Longchain can list collections, retrieve documents as JSON, inspect schemas, run… | ||
| Software Design: Humans vs AI | Olivier Azeau | 43m |
A live classroom game called "Software Ball" runs on phones to measure code cost by having players pick simple versus generic snippets and combine them to meet… | ||
| My entire stack is in my Postgres DB and a few power tools for my backend | Loïc des Rochettes, Mathieu Passenaud | 47m |
A backend architecture that keeps the entire stack inside PostgreSQL: schemas define models and views, PostgREST exposes the views as a REST API, and an… | ||
| AI and Edge Computing: Efficient, Lean, Dependency‑Free Language Processing | Didier Plaindoux, Ionut Mihalcea | 47m |
The talk compares two approaches for doing language processing on-device without external services: self-hosting large open-weight generative models (targeting… | ||
| Barcodes, it's more than just bars | Sylvain Gougouzian | 45m |
History and facts about barcode evolution from the 1952 UPC patent and the first supermarket scan in 1974 to laser scanner hardware using galvanometric mirrors… | ||
| Content Design Saves Your Forms | Nathalie Rosenberg | 46m |
Argues for a content-first workflow instead of lorem ipsum and shows how applying four Nielsen heuristics (consistency, match-to-world, error prevention, clear… | ||
| Observability in Practice: How to Become Proactive with Bugs | Mathieu Lamiot | 49m |
Practical observability techniques using Sentry or GlitchTip for full error context (stacks, variables, occurrence counts), Metabase SQL-based data alerts that… | ||
| A Business Backend for My IDE | Aless Hosry, Benoit Verhaeghe | 37m |
An LSP-based backend built in Pharo exposes model-driven analyses to IDEs by combining Moose Technology (Famix/Fast metamodels), the Motion pattern language… | ||
| Beautifying QR Codes in the Age of GenAI | Raphaël Semeteys | 16m |
A pipeline that converts plain QR images into decorative yet scannable art using QR Tool Kit for generation and analysis, ComfyUI with Stable Diffusion 1.5 for… | ||
| Increasing Your App's Accessibility to Optimize It for All Users | Kevin Abrioux | 40m |
The talk explains how to make an iOS app accessible and compliant with the EU Accessibility Act (effective 28 June 2025) using Swift and Xcode. It covers… | ||
| Rewrite It In Rust: Can We RIIR Everything? | Igor Laborie | 42m |
Argues that rewriting unsafe C/C++ components in Rust removes most memory-safety vulnerabilities while keeping C-like performance without a garbage collector… | ||
| I Abuse Your LLM Live with 10 Security Vulnerabilities | Gaëtan Eleouet | 44m |
A live demo shows ten concrete failure modes of large language models, including prompt injection that reveals hidden system prompts, agent tool calls that… | ||
| Open Sesame Secure your SSH server with Port Knocking | Paul Jeannot | 16m |
Demonstrates securing an exposed SSH server on Debian 12 by using iptables and the port-knocking daemon Nockd to keep port 22 closed until a specific sequence… | ||
| Chance Makes Good Tests | David Pilato | 14m |
A bug-hunting workflow for Java unit tests using a randomized-testing framework that injects and logs the exact random seed so failures can be reproduced. The… | ||
| Reactivity and signals demystifying frontend magic | Estéban Soubiran | 42m |
Presents how the Alien Sinal library implements reactive signals, computed values and effects using a push/push‑poll hybrid that flags parent nodes while… | ||
| From chaos to elegance refactoring complex workflows with events | Thibaut Cantet | 43m |
Refactors a messy Roland Garros ticket payment flow (PSP, 3DS, backoffice) into a synchronous event-driven choreography that breaks monolithic use cases into… | ||
| Student and developer Bouncing back after failures in creating Fillzz | Thomas Roux | 18m |
A student-built mobile app ingests the data.gouv.fr dataset to track fuel prices (about 8 million prices/day across 271,000 stations in 20 countries) and uses… | ||
| JEP 493 this new Java feature is validated | Guillaume Benoot | 15m |
Describes the JEP (JDK Enhancement Proposal) process that turns ideas into Java code: a JEP is required for any substantial change (work over two weeks) and… | ||
| nothing matches | ||