IA
Conférence
Horaire à confirmer
How can developers use AI to explore an unfamiliar domain such as game development? This session will follow a small game-development experiment using C#, .NET, Visual Studio Code, Unity, and accessible cloud-based AI tools. The focus will be on the development process: defining a specification, guiding an AI agent, shaping the architecture, reviewing generated code, and validating the result. Along the way, we will connect these practices to web development and other areas of software engineering. No prior game-development experience is required; the session is an exploration of how AI can help developers learn, experiment, and build beyond their usual areas of expertise.
IA
Conférence
Horaire à confirmer
I built a digital microscope that combined IoT and AI into something useful, practical, and centred on accessibility. It could see, read, and talk, and it ran entirely in the cloud. This year, the journey is moving all of it from the cloud to the edge.
Join me as we take every capability off the cloud and onto the device itself. Vision, text recognition, speech, conversational AI, and even retrieval, all shifting from hosted services to models running locally on a small device, with no connection required. We will explore what each move from cloud to edge costs, where running locally surprises you, and what living on the device makes possible that the cloud never could.
Best of all, this is all still done with .NET and Avalonia.
Going from cloud to edge means this accessibility solution now travels with me, signal or not. Hopefully the session sparks some ideas about what else becomes possible when you move the intelligence onto the device.
IA
Conférence
Horaire à confirmer
Enterprise-grade RAG: guardrails, semantic caching and security.
IA
Conférence
Horaire à confirmer
Building AI applications usually means relying heavily on cloud-hosted models, which introduces high token costs (COGs), unpredictable latency, and strict data privacy boundaries. While tools like Ollama or LM Studio offer local alternatives, they require your end-users to install heavy, external background dependencies, complicating the app distribution and installation pipeline
Enter Microsoft Foundry Local, a lightweight, 20-megabyte, on-device AI runtime that solves these challenges. It packages a highly optimised local AI stack directly inside your process, eliminating the need for self-hosting setups, Docker containers, or external software installations
In this session, we will deep-dive into how Foundry Local leverages the performance of ONNX Runtime and WinML to automatically select and run the absolute best variant of a model for your user's specific hardware, whether they have an NVIDIA GPU, Intel/Qualcomm NPU, Apple Silicon, or a standard CPU
IA
Conférence
Horaire à confirmer
Modern agentic AI systems rely on many moving parts, from IChatClient pipelines to Microsoft Agent Framework agents to MCP servers that expose external tools and capabilities. As these components interact, the need for reliable, deterministic testing becomes critical.
You need to ensure:
- Agents reason correctly and follow their execution loop
- Tools are discovered and invoked with the right parameters
- Responses stay semantically accurate across iterations
- Multi‑agent and multi‑server orchestration behaves predictably
This is exactly where skUnit excels, a robust .NET testing framework purpose‑built for agentic AI units, including IChatClient, Microsoft Agent Framework agents, and MCP integrations. Whether you're validating a single agent’s reasoning step or end‑to‑end orchestration across multiple tool surfaces, skUnit gives you production‑grade confidence in your AI systems.
IA
Conférence
Horaire à confirmer
Microsoft Agent Framework is Microsoft’s cutting-edge platform for building intelligent, autonomous, and collaborative AI agents. It enables developers to create agents that can reason, communicate, and integrate seamlessly with enterprise systems. In this session, we’ll explore its architecture, key features, real-world applications, and how you can start building your own AI-powered solutions today.