Dev Summit 2026

What if your best SRE was an AI capable of analyzing millions of signals in seconds?

In this session, I’ll show you how we built a "Home-Made" Incident Investigator to automate root cause analysis from the ground up. By leveraging the Model Context Protocol (MCP) to unify Datadog telemetry, GitHub source code, and ServiceNow tickets, our autonomous agent doesn’t just report issues—it investigates them.

We will dive into how the agent iteratively forms hypotheses, "peeks" into raw logs, and suggests code fixes the moment an alert fires. No more cognitive overload or endless manual digging during high-pressure outages. Come and learn how we are drastically slashing our MTTR (Mean Time To Recovery) by adopting an "Agent-First" approach to production operations.

Modern cloud-native platforms have dramatically improved how we build and deploy software.
Yet many organizations still struggle with fragmented releases, inconsistent promotions, limited traceability, and unclear ownership between CI, CD, and Operations.

This talk explores how Release Management can evolve from a technical mechanism into a true product capability.

We will cover:

• The gap between CI/CD tooling and real Release Management needs
• Common anti-patterns (pipeline-driven releases, manual promotions, hidden drift)
• The Digital Component approach
• Versioning, promotion flows, and rollback strategies
• The role of observability and governance
• How product thinking transforms Release Management adoption

Through practical examples from a large-scale platform environment, we’ll show how to move from “deployment success” to “controlled value delivery.”

If you think UI Design is still about moving a button 1 pixel to the left, I have some bad news: those pixels are dead. But here is the great news: Design has never been more alive.

In this keynote, we’ll explore the paradigm shift from "pixel pushing" to "Design orchestration." We will dive into how AI is transforming the UI Designer's role : moving from manual execution to strategic direction. Using real-world examples from Adeo’s CCDP, we'll see why the automation of our craft isn't a threat, but a liberation. Finally, we will discuss the "Human Moat": why empathy, ethics, and "habitant intuition" are the only things an algorithm can't simulate.

Join me to discover why the future of UI isn't about teaching machines to design, but about using machines to finally give us the time to be a designer.

As part of a migration project from a legacy, in-house Product Lifecycle Management (PACE) solution, using Copilot helps us accelerate documentation, specification reprocessing, and also develop and test faster.

This talk will allow us to share our learnings and the concrete results we have achieved.

Summary (in progress):
- Context of the migration
- Problems and deadlines
- First try
- Don't do it on your own
- Aim for the moon (trying to scale)
- Setting up a new way to work
- Concrete results
- Learnings

Digital transformation often focuses on building the right product, but success actually depends on adoption at scale.
At ADEO, with more than 120,000 collaborators across multiple countries and business units, launching a platform or product is only the beginning. Ensuring that solutions are understood, adopted, and embedded into daily operations requires a structured Change Operating Model.
In this talk, I will share how ADEO designed a “Change by Design” approach, integrating change management directly into the digital framework and delivery lifecycle.
We will explore how platforms, product teams, and business units collaborate to transform delivery into real impact, with the ultimate ambition of evolving from Change by Design to Change by Default.

This presentation covers the evolution of Supply Chain automation at Leroy Merlin Ukraine. How the decommissioning of our old version of Pyxis led us to an AI Agent-enabled workflows in *almost* one step.

- The Challenge: Balancing legacy system deprecation and an upcoming SAP migration while trying to eliminate the manual resolution of micro-errors (e.g., typos, conflicting addresses) that traditional automation cannot catch.

- The Breakthrough: Proving that AI Agents succeed where rigid algorithms fail. By mimicking human reasoning to clean up unpredictable user inputs, this architecture is designed to directly reduce our Manual Touch Rate (Exception Rate) and drive up our First-Attempt Delivery Rate (FADR).

- The Demonstration: A practical look at our "start small, explore patterns, test assumptions" philosophy, showing how AI handles the tedious exceptions so humans can focus on delivering value.

This is a perfect opportunity to address smaller BUs like ourselves with a viable, scaled-down approach to Agentic AI. The goal: eliminate the countless micro-errors that constantly drain Supply Chain teams already operating at full capacity.

Diversity is often discussed in broad strokes, but what does it look like on the ground, in the middle of a noisy open space, when standard communication fails?

In this session, I will share the honest, unfiltered story of my collaboration with Hatem. When Hatem first joined as a contractor, the challenges seemed insurmountable: no eye contact, a total reliance on a job coach to communicate, and the sensory overload of a bustling tech environment. Fast forward two years, and Hatem is not just a permanent employee—he is a pillar of the team and a recognized expert across the organization.

We will walk through the heavy initial investment required to build trust and the specific adjustments—from daily 1-on-1s to "silent rooms"—that turned a difficult start into a technical breakthrough. You will discover how Hatem transitioned from a QA role to a QE (Quality Engineering) expert, mastering automation with the help of a structured mentorship model.

Key Takeaways:

- Beyond Soft Skills: How to adapt management for profiles that don't fit the "standard" social mold.

- The Mentorship Blueprint: Why the "Show once, let fly" approach works wonders for neurodivergent talent.

- Creating the Right Environment: Practical tips on handling sensory challenges (noise, open spaces) in tech offices.

- The ROI of Patience: Why investing time in neurodiversity leads to high-autonomy, high-impact technical experts.

In 2021, Copilot helped me finish lines of code. By 2024, I was chatting with it. In 2025, Context Engineering was the breakthrough. Now in 2026, I don't just talk to AI — I delegate to it.

Copilot CLI, Skills, agentic workflows, new models — I've gone from asking for help to handing over entire tasks, from scaffolding projects to opening pull requests, all in natural language. And it's not just me: at GitHub, our own product teams are using these same capabilities to ship faster and rethink how software gets built.

The shift goes beyond developers. Product managers, SREs, DevOps engineers — they're all becoming builders. It's no longer about writing code, it's about expressing intent.

In this talk, I'll show you what changed between 2025 and 2026, and how it turned me from a developer with a great assistant into a complete builder.


Note: In this session I will focus on the new features and primitives that have change my life, and GitHub's developers life too. Obviously the CLI, SDK, Agentic Workflows, .. is a big change, but also some features that we are launching in few weeks (I am not allowed to talk about it yet), that is helping, developers but also people in Product Management, Marketing, ... (The presentation will be a mixt of feature introductions, demonstration and stories from our team.

Is your domain code becoming complex, hard to maintain, and disconnected from real business needs? You keep hearing about **Value Objects**, **Aggregates**, and **Entities** at conferences but struggle to see how to apply them concretely?
This 1.5‑hour hands‑on workshop will demystify **Domain‑Driven Design (DDD)**. Forget abstract diagrams! We will start from a simple, collaborative business exploration (inspired by Event Storming) to let DDD tactical building blocks emerge naturally.
Together, we will incrementally code a small business domain in __Java__. Through **practice**, you will see which concrete problems a Value Object or an Entity solves, how the Aggregate safeguards data consistency.
Whether you are a curious developer or looking to better structure your applications, this workshop will give you the keys to move from theory to practice and start applying DDD in your projects as of tomorrow.

With the advent of the age of AI software has never played such an outsized role in our daily lives and yet women only account for a small minority of software engineers. This is especially disheartening given that many - but not all - of the first software engineers were women (Ada Lovelace, Katherine Johnson, Margaret Hamilton, Grace Hopper, ...).

This talk aims to stay away from dogma or political posturing, instead focusing on explaining why more women aren't software engineers and proposing concrete, tactical and doable solutions.

The first part of the talk recounts the stories of some of the pioneers who built flight control systems for the Apollo mission, lamenting that their stories aren't more widely known.

I then use those examples to contrast with the current state in our Industry where female developers only account for a minority of software engineers.

I'll then give an overview of why I believe more women don't become software engineers, briefly explaining why this is the case.

I'll also share why some of the women who become software engineers eventually leave our Industry and what we can do to retain them.

Finally, I'll conclude with concrete actions we can all take in order to increase the amount of women in software engineering.

This is a tentative/more detailed outline of the talk:

Introduction: the pioneers.
Ask attendees to look around them, notice how little women are in the room.

This wasn't always the case and we have many examples of women who've made massive contributions to software engineering: from Ada Lovelace to Grace Hopper.

How we got here.
The first part of the talk uses a data driven approach to describing the problem at hand, focusing on how many women are developers.

Why we got here: women in STEM
After underlining the fact that education is key and that only a small amount of women pursue an education in STEM I'll briefly explain why that is.

Adding insult to injury: female software engineers in the workplace
I'll then continue on by explaining that even when they are software engineers a lot of women end up leaving our Industry despite haven chosen a career path in software engineering.

What can be done about it?
Finally, I'll focus on the possible solutions we as individuals can implement in order to increase the amount of women who become software engineers and make our teams more diverse. I'll focus on the power of mentorship, the importance of retaining the female software engineers on your team by fostering a culture of respect, empathy and excellence.

AI and software are transforming our lives and shaping how we view, interact with and perceive the world; and yet software engineering teams are way too often stacked with men. This talk isn't about political posturing or gimmicks, it's about explaining how we got here, providing data on what the current state of affairs is, why this problem matters (importance of building diverse teams) and what can be done to solve it. The solutions proposed don't include quotas or political dogma but instead focus on concrete solutions like building role models, fostering mentorship and encouraging girls to pursue STEM education.

Are you a Product Manager looking to maximize your product's impact? Do you want to understand how Product Marketing fits into a Product Operating Model?
Product Marketing is a strategic partner throughout the entire product lifecycle. Join us for a deep dive into the synergy between Product and Marketing, featuring real-world applications.

Management decisions about engineering teams too often rely on incomplete signals — sprint velocity, subjective updates, and what surfaces in status meetings. But the data to make better decisions already exists: in your Jira and GitHub, waiting to be connected.

We used AI-assisted analysis to extract flow metrics from Jira and DORA indicators from GitHub — and what emerged was a picture no one had seen before. Trends in change failure rate, rework patterns, and team load revealed systemic issues that weren't showing up in any dashboard or review meeting. The data didn't just confirm suspicions — it reframed the conversation entirely.

What followed was a structured change process: workshops grounded in evidence, renegotiated team agreements (Definition of Ready), and a shift in engineering practices — driven not by management mandate, but by shared understanding of what the numbers were saying.

This talk is for managers who lead engineering teams and want to move from intuition to insight.
You'll learn:

- How to surface flow metrics from Jira and DORA indicators from GitHub
- What patterns signal team stress, quality erosion, and instability — before they become crises
- How AI lowers the barrier to analysis: no special skills required
- How to use metrics to drive change conversations — not performance evaluations

The experiment is still in progress. We'll share early signals, honest lessons, and what we'd do differently.

At conferences, we usually showcase our best products with elegant interfaces, smooth user journeys and impressive adoption rates.
But if we're honest, digital products are still making the same design mistakes again and again.

I want us to remind together the worst of UX: confused users, friction points at every steps, leading us to wrong decisions. Invisible actions, misleading buttons, dreadful forms,... We all know that this is bad so why do we still keep seeing them online?

Let's explore through funny examples, the strange, the surprising or deceptful designs that made us yell altogether "Bad UX!"
Because good experience is most of the time not about innovative patterns but about preventing and avoiding obvious mistakes, we can questions ourselves about quality of the experience we want to deliver to our users.

Ever since vibe-coding hit the scene, you might think knowing how to code is old news for building apps thanks to AI. But let's be real, some of us who dove in got burned: hallucinations, performance and security issues, messy code, and the final product not quite matching the initial idea.

What if there was a way to **get the best of both worlds**? To boost efficiency leveraging AI while being 100% confident when pushing to production?

We chose **Spec-Driven Development** using **SpecKit** to build an application we deployed with total confidence, using **English** (and some French\!) as our main "programming" language\!

In this talk, Christine (Product Manager) and Julien (Staff Software Engineer) will reveal:

* **The "state of the art"** in this space and where the industry stands today.
* **What is Spec-Driven-Development**, in a nutshell
* How AI helped us build the *right* software by supporting **both the PM and the developer**.
* **Ensuring code quality** while dodging nasty surprises.
* **Real-world struggles** and the pitfalls you’ll want to watch out for.
* **Where we go from here** \- the areas we're still refining and what we've yet to solve.

Join us to learn how to move beyond the hype and embrace a development process that we found smart, effective, and production-ready\!

Artificial Intelligence is often seen as a virtual, weightless force. Yet, its environmental footprint—from GPU intensive training to daily inference—is very much physical. As Decathlon and Adeo accelerate their AI integration, a crucial question arises: Can our innovation be truly "forward-thinking" if it isn't sustainable?

In this session, I will take you behind the scenes of a 2025 Proof of Concept that bridges the gap between raw compute power and ecological responsibility. We will explore how leveraging cloud environmental telemetry allowed us to establish the first reliable & actionable metrics for AI environmental impacts.

Key takeaways from this talk:
- The Foundation: How cloud-native monitoring tools serve as the "entry point" for measuring AI impact.
- The 2025 POC: Real-world feedback on integrating environmental metrics into the AI lifecycle.
- Actionable Strategies: Moving from "Carbon Awareness" to "Concrete Environmental impact limitation" without sacrificing model performance.
Let’s stop treats AI as a black box and start building an architecture that respects both our business goals and our planet.
Call to Action & Synergies Decathlon + Adeo

Imagine if we could understand a user’s intent while they are navigating.Not just what they click but what they are trying to achieve.
Are they looking for inspiration? Designing a project?Are they exploring broadly, or focused on a very specific product universe?
If we could detect that in real time, we could adapt the experience instantly, surfacing the right contents and products among the millions we have.
That’s the promise of clickstream: turning user behavior into signals of intent for a continuously experience improvement

In this session, we will present how clickstream captures and streams millions of user navigation events per day in milliseconds timeframe and how these events are enriched and transformed into meaningful signals and aggregates to empower user journey.

Based on case studies and product demo, we will explore how to:
Manage at scale an intensive volume of events in real time processing
Enrich events on the fly (case with Knowledge graph data)
Create aggregates & signals using Kafka streams
Connect features on Clickstream.

Attendees will leave with a clear understanding of real time data processing and opportunities behind intent signals detection capabilities