CI/CD That Understands Intent: Continuous AI with Agentic Workflows

CI/CD has mastered deterministic checks—linting, tests, SAST, schema validation—but modern engineering risks increasingly live in the nuanced, contextual layer: “Does this change violate privacy expectations?”, “Did the UX drift from the spec?”, “Is this workflow introducing risky behavior?”

In this talk, we’ll explore non-deterministic CI gates: AI-assisted checks that evaluate context, intent, and policy, not just fixed rules. We’ll show how GitHub Agentic Workflows bring “Continuous AI” into your existing GitHub Actions pipelines by authoring automation in plain Markdown, executed by coding agents with security-first guardrails.

You’ll see practical SDLC examples—like policy/compliance analysis and UX alignment gates—where agents produce human-reviewable, structured findings and can safely trigger pre-approved operations via controlled outputs.

Walk away with a blueprint for augmenting deterministic CI/CD with agentic gates to drive faster delivery, reduced risk, improved developer experience, and smarter automation—without sacrificing governance.

From Vibe Coding to Vibe Engineering: Building Reliable AI-Driven Development with GitHub Copilot

Vibe Coding has shown developers how quickly ideas can turn into working code with GitHub Copilot. But as teams move beyond experimentation and into production, speed alone is no longer enough.

In this demo-heavy session, we explore the evolution from Vibe Coding to Vibe Engineering—where AI agents are guided, constrained, and standardized to deliver high-quality, repeatable outcomes. You will see how GitHub Copilot’s agentic primitives—including Custom Agents, Instructions, Prompts, Skills, Hooks, and Plugins—enable teams to embed best practices, guardrails, and quality controls directly into their development workflows.

Through live demonstrations and real-world examples from enterprise teams, we’ll show how these primitives allow developers to move from ad‑hoc prompting to intentional agent orchestration, ensuring consistency, security, and engineering rigor at scale.

We will cover practical patterns, anti‑patterns, and lessons learned from the field to help you confidently apply agentic development in your own organization.

Model-Care-Coach: Foundation of leadership at Microsoft

In today’s dynamic work environment, effective leadership transcends traditional boundaries. Microsoft’s Model-Care-Coach framework embodies this evolution, emphasizing three pivotal dimensions that empower leaders to thrive in turbulent times.

In this talk, we’ll explore each one of the core components in detail and delve into how the Model-Care-Coach model aligns with Microsoft’s growth mindset, fostering resilient leaders who navigate change with compassion and purpose.

During my time at Microsoft, I experienced and learned about the Model-Care-Coach leadership approach and its benefits: It was eye-opening! I was finally able to find a management style I would identify with.

Join me to learn about why this model is such a foundational pillar of leadership at Microsoft.

Agent Package Manager (APM) - dependency manager for AI Primitives.

In this workshop, we explore a growing challenge in the era of AI‑assisted software development: AI coding agents are only as effective as the context they’re given — and today, that context is scattered, manual, and inconsistent across teams.

Prompts, skills, instructions, plugins, and coding standards often live in local folders, shared Slack threads, or personal setup scripts — making agent behavior difficult to reproduce, review, or scale across environments. As a result, developers working on the same repository can end up with vastly different AI‑assisted experiences depending on how their agent is configured.

We will introduce the Agent Package Manager (APM) — an open‑source dependency manager designed to bring software engineering discipline to AI agents by treating agent configuration as a versioned, installable dependency

Just like npm, nuget or pip for application code, APM resolves transitive dependencies, pins versions for reproducibility, and installs all required agent primitives into the project — ensuring consistent AI behavior across developers, CI pipelines, and environments