Web AI: Smarter Web Experiences Directly in the Browser

Ever thought you could run AI right in the browser, with zero data leaving your users’ devices? Thanks to on-device AI technologies and modern hardware acceleration, that future is here. In this talk, we’ll break down the trade-offs between client-side and server-side AI—helping you balance performance, privacy, and responsiveness. We’ll explore tools like Chrome’s built-in AI APIs, Transformers.js, WebLLM, and Ollama, showing you how to build smarter apps right at the user’s fingertips. Get ready to unlock a new era of web development, where AI isn’t just on the backend—it’s right there in your browser, delivering instant, interactive intelligence.

Web AI: browser-powered intelligence

Artificial Intelligence (AI) is revolutionizing the web, but integrating AI models into websites has been challenging due to their massive size and the complexity of deployment. In this presentation, we'll discuss the advantages and limitations of client-side versus server-side AI. We'll dive into Web AI technologies, including Mediapipe, TensorFlow.js, ONNX Runtime Web, Transformers.js, WebLLM and Chrome's built-in AI. Attend the presentation and learn skills to bring Web AI magic into your own projects. No servers, no API keys, no usage limits — just raw power of AI in the browser.

Context-aware Agents into Web Apps

In a galaxy not so far away, every user deserves a trusty sidekick like R2-D2—a copilot that truly understands their needs. What if our web apps could offer the same seamless interaction? In this presentation, we'll explore how to embed AI Agents into web apps to read contextual data, such as user inputs, form elements, and navigation patterns. These copilots can make your app smarter and responding to context in real-time. May the AI Be with you!

What you will learn:

- Embed a contextual chat feature: Engage users by answering queries and performing tasks based on conversational inputs.
- Enable action-driven assistance: Allow the copilot to autonomously perform tasks, boosting interactivity and efficiency.
- Offer dynamic content recommendations

AI Agents in web apps

In a galaxy not so far away, every user deserves a trusty sidekick like R2-D2—a copilot that truly understands their needs. What if our web apps could offer the same seamless interaction? In this presentation, we'll explore how to embed AI copilots into NextJS apps to read contextual data, such as user inputs, form elements, and navigation patterns. These copilots can make your app smarter and responding to context in real-time. May the AI Be with you!

What you will learn:

- Embed a contextual chat feature: Engage users by answering queries and performing tasks based on conversational inputs.
- Enable action-driven assistance: Allow the copilot to autonomously perform tasks, boosting interactivity and efficiency.
- Offer dynamic content recommendations: Suggest personalized content, tailoring the user experience to individual needs.

Extracting Data Insights from Documents, Images, Audio, and Video

Most organizations’ best insights live outside neat tables—buried in PDFs, scans, screenshots, call recordings, and long-form video. This talk shows how to turn that unstructured mess into clean, trustworthy JSON you can ship to search, analytics, and automation. We’ll build an end-to-end extraction playbook in C# using Azure AI Content Understanding’s Analyzer to ingest multimodal files and return structured outputs via a simple POST → GET pattern. We’ll compare when to use Azure Content Understanding, Azure Document Intelligence versus rolling your own with Azure OpenAI. Expect live examples. This session turns documents into assets you can query tomorrow.

You’ll learn
- Detect → segment → extract → validate → persist pipeline, with multimodal tactics for each step.
- When to use Document Intelligence vs. Content Understanding vs. Azure OpenAI—trade-offs in accuracy, flexibility, and ops.
- Production guardrails: confidence thresholds, schema/regex checks, PII redaction, and eval metrics.

Who should attend
Engineers, data practitioners, and architects shipping real pipelines—not just proofs of concept.

From Unstructured Data into Structured: Azure AI Content Understanding in Practice

Most of your valuable info isn’t in tables—it’s stuck in PDFs, images, audio, and video. In this session, we’ll use Azure AI Content Understanding to turn that unstructured content into clean, structured data your apps can actually use. We’ll walk through the core Analyzer flow, then show real examples: extract fields from invoices, summarize calls, and segment videos into searchable scenes. We’ll also compare when to choose Content Understanding vs Document Intelligence vs a DIY Azure OpenAI pipeline. We’ll finish with a small, production-minded template you can use. By the end, you’ll know how to unlock hidden data across formats and wire it into search, analytics, and automation.

What you’ll learn
How to extract entities/summaries from documents, audio, and video
Decision guide: Content Understanding vs Document Intelligence vs DIY
Production basics: schemas, PII redaction, confidence gates, logging

On being a good Copilot Coach

With the advancement in Generative AI and Large Language Models, there's a fundamental shift on how we develop applications. Software engineers are not just building blueprints and code instructions, they are being called to be good Copilot coaches. Attend this presentation and learn practical strategies to effectively coach Agentic AI systems, allowing it to find the better version of itself.

Azure AI for Document Processing: What to Use When

Developers face two solid options on Azure: Document Intelligence for structured/semi-structured docs, and Content Understanding for multimodal workloads (docs, images, audio, video). In this session we compare capabilities, cost/risk trade-offs, and when to choose each. We’ll build a small pipeline that extracts fields from invoices with Document Intelligence, then do the task with Content Understanding.

On-Device Generative AI: Intelligent Apps, No Cloud Required

Forget waiting on the cloud—AI is now running directly on your devices! This shift to on-device AI transforms how we interact with smartphones, tablets, desktops, and even web browsers, offering instant results, better privacy, and smoother user experiences.

In this practical session, you'll see real-world demos in C# and JavaScript, showcasing how to integrate generative AI locally into web, mobile, and desktop applications. We'll explore built-in AI capabilities in Chrome, Windows, and mobile apps, highlighting when on-device AI is the ideal choice—and exactly how to implement it in your next project.

Human-Agent Communication Challenges

Sometimes AI feels like it can read your mind—until it doesn’t. Small gaps in what you meant (“cheap, but not a red-eye”) or what the AI assumed (“any cheap flight is fine”) turn into big problems once agents can act: editing files, booking trips, or moving money. If an agent misunderstands you—or gets tricked by a malicious web page (prompt injection)—it can buy the wrong ticket, leak info, or mess up your stuff. The risk isn’t just bad answers; it’s unwanted actions.

This talk turns the research on Challenges in Human-Agent Communication into a simple, practical checklist you can use right away. We’ll focus on four clear questions for every agent task:
Can-do (What are its abilities and limits?) → Plan (What will it do next?) → Doing (What is it changing right now?) → Done (What actually happened, with proof?). You’ll see concrete patterns like: show capabilities up front, preview with a diff before committing, use scoped permissions and approvals, attach evidence for results, and ask for help when uncertain.

AI in the Browser: Real-Time ML

AI doesn’t have to mean Python, servers, or user data leaving the device. We’ll run models in the browser with TensorFlow.js, Transformers.js, and modern Web APIs—no installs, instant demos. You’ll see real-time tasks like vision, audio, and small LLMs distilled for client-side use, plus practical limits and workarounds. We’ll end with a tiny, shippable demo pattern you can reuse. By the end, you’ll know when to go on-device vs. cloud, how to wire JS inference, and how to talk about privacy/perf trade-offs with your team.

Agentic AI Patterns in C#: Plan → Act → Evaluate

Agents shouldn’t feel like magic—they should feel testable. In this session you’ll learn the core patterns behind agentic systems in C#: planning, tool use, memory, reflection, and evaluation loops. We’ll implement a small, real-world workflow (e.g., “triage → decide → perform action → verify”) using .NET and common AI SDKs, then add the guardrails that turn a demo into something you can ship: schema-based tool calls, scoped permissions, retry/rollback, and behavior evals in CI.

To make it stick, we’ll use a quick Minions metaphor—specialized “little workers” that each do one job well—but we’ll keep the focus on production code. You’ll leave with copy-and-pasteable patterns, a reference repo, and a mental model to scale from single-agent helpers to multi-agent teams without chaos.

Agentic AI Orchestration Patterns in C#

Big problems get messy when one supersized AI tries to handle everything. Break the job into bite-sized chunks and hand each piece to a focused agent—keeping prompts and code short, readable, and easy to debug. Need a new skill? Just plug in another agent instead of rebuilding the whole stack. Join us to explore AI Agent Orchestration Patterns — ready-to-use playbooks that show your agents how to chat, trade work, and crush complex missions together.

AI Agents on Demand

What do James Bond, Ethan Hunt (Mission: Impossible) & Jason Bourne have in common? They’re on-demand agents — hired for a mission, bringing their skills, tools, and vanishing after the job’s done. AI Agents work the same way: purpose-built, on-demand, and mission-ready. Your mission, should you choose to accept, is to learn how to build scalable AI agents with Azure AI Agent Service in .NET — combining reasoning, tools & automation for next-gen cloud solutions. This message will self-destruct in 5 seconds.

Context-Aware Copilots for Next.js: Read UI, Decide, Act (Safely)

In this session, we’ll embed a context-aware copilot into a Next.js app that actually understands what’s on screen—form values, selections, and route state—and takes safe actions your users approve. We’ll wire a conversational UI to real capabilities via typed tool calls, gate risky actions with consent prompts, and show diffs/undo before anything commits. You’ll see how to pass front-end signals to the model, decide when to chat vs. call tools vs. retrieve content, and recover gracefully when confidence is low. We’ll close with a small template you can reuse in your app (context → decide → act → audit). By the end, you’ll know how to add an R2-D2-style sidekick to your product—helpful, legible, and secure. May the AI be with you.

What you’ll learn

- Capture UI context (forms, selections, route params) safely with a “context bridge.”
- Expose actions as typed tools with consent, diff/undo, and audit logs.
- Route requests (chat → tools/RAG) using simple confidence gates.
- Generate in-page recommendations tied to the user’s current task.

Who it’s for:
Web/Next.js engineers; basic React + LLM API familiarity.

Mixed Reality for web developers

Mixed Reality (MR) has entered the web space with WebXR, enabling immersive experiences directly through browsers like Edge, Chrome, and the Meta Quest browser. Dive into this session to harness JavaScript for crafting captivating MR applications. Explore top-notch open-source AR/VR JavaScript libraries to kickstart your projects. Get ready for a hands-on demo that brings the potential of Mixed Reality to life. Join us to push the boundaries of digital interaction and step into the future of web-based MR.

Ship Agentic AI on Azure: Tools, Plans, and Guardrails

Modern apps need agents that can reason, call tools, and get work done—safely. In this session, we’ll design a small, production-minded agent: define its mission, give it tools (APIs, retrieval), add planning, and wrap it with guardrails (scopes, approvals, logging). You’ll see code for orchestrating tool use, handling uncertainty with confidence thresholds, and recovering when plans go off-track. We’ll also discuss where agents shine, where they don’t, and how to measure usefulness beyond “wow” demos. By the end, you’ll have a repeatable pattern to stand up task-focused agents on Azure, plus a checklist for safety, evaluation, and cost control.

What you’ll learn:
- Model a task-focused agent (mission, tools, planner, memory).
- Add guardrails: scopes, approvals, audit logs, fallbacks.
- Evaluate behavior and track quality over time.

Who it’s for: NET engineers shipping real features, not just demos.

Web AI: browser-powered intelligence

Artificial Intelligence (AI) is revolutionizing the web, but integrating AI models into websites has been challenging due to their massive size and the complexity of deployment. In this presentation, we'll discuss the advantages and limitations of client-side versus server-side AI. We'll dive into Web AI technologies, including Mediapipe, TensorFlow.js, ONNX Runtime Web, Transformers.js, WebLLM and Chrome's built-in AI. Attend the presentation and learn skills to bring Web AI magic into your own projects. No servers, no API keys, no usage limits — just raw power of AI in the browser.

Why AI in JS

Many web developers eager to dive into AI hit a common roadblock: they think learning Python is the only way to start. This talk breaks that myth and shows when AI in JavaScript makes perfect sense—and when Python still shines. AI with JavaScript offers a friendly, accessible entry point that doesn’t require deep math skills or heavy frameworks. Join me to explore live demos featuring TensorFlow.js, ONNX Runtime Web, MediaPipe, Transformers.js, and Chrome’s built-in AI—all running right inside the browser. Why choose AI in JS? Because it means zero installs, better client privacy, and true cross-platform power. Say goodbye to Python-only pipelines and heavy servers—embrace the future of AI on the web!

Why AI in JS

Most developers would agree that getting started with AI development can be intimidating. AI with JavaScript provides an approachable entry point, allowing devs to dip into AI without needing to master complex math or frameworks. Attend this session and see code demos running Tensorflow.js, Mediapipe, Transformers.js, Chrome Built-in AI - all running inside of the browser. Why? To ditch python for AI. Heck, ditch servers for AI too! AI in JS brings Zero installs, Client privacy, and Cross-platform support.

Developing Spidey Sense : Anomaly Detection for apps

Anomaly detection is the process of identifying unexpected items or events in data sets. It’s about detecting the deviation from expected pattern of a dataset. It’s like having “spidey senses” for your apps that can detect when there’s danger or something is not right. Attend this session and learn about using anomaly detection in ML.NET, Azure Stream Analytics and Cognitive Services API; become a superhero and save the day.

Agentic AI Design Patterns

As artificial intelligence continues to evolve, we're witnessing a paradigm shift in how AI systems are structured and deployed. Instead of monolithic models, modern AI architectures increasingly rely on multiple specialized agents working together to accomplish complex tasks. In this presentation, we’ll explore Agentic AI design patterns by drawing a playful comparison to the minions. Just like these quirky characters, AI Agents autonomously perform tasks, follow instructions, and work together toward a common goal. Through this lighthearted analogy, we’ll see how minions' chaotic teamwork mirrors the structured collaboration of AI agents, helping you grasp these advanced concepts in a fun and relatable way.

Predicting the future (of equipment) using ML.NET

Predicting the future is not easy! Forecasting peak operating conditions and avoiding unexpected equipment breakdown is possible, that’s the goal of Predictive Maintenance. Attend this session and learn how to train a machine learning model in Visual Studio with ML.NET Model Builder, and build an app that uses the model. Fair warning - I cannot predict how live demos would go. I will do my best.

AI Coaching Playbook: Strategies for Multi-Agent Systems

With the advancement in Generative AI and Large Language Models, there's a fundamental shift on how we develop applications. Engineers are not just building blueprints and code instructions, we are being called to be good AI coaches. Expert Coaches that can guide multiple AI agents to discover, learn, collaborate and/or compete with each other. Come and join the presentation to learn practical strategies to effectively coach AI systems, ensuring they perform at their best and allowing us to find the best version FTW!

Managing IoT Devices in Azure; the simpler way

Collecting telemetry data from devices is a common requirement today for any IoT solution. In this session, we will learn about Azure IoT Central and how it reduces the burden and cost of developing, managing, and maintaining IoT applications. Journey with me as we will walk through different features of IoT Central: connect devices, monitor, configure rules, and integrate with other services. Learn how to customize IoT central to fit the needs.

Making neural networks portable with ONNX

The world of machine learning frameworks is complex. What if we can use the lightest framework for inferencing on edge devices? That’s the idea behind ONNX format. Attend this session and find out how to train models using the framework of your choice, save or convert models into ONNX, and deploy to cloud and edge using a high-performance runtime in python, C# or javascript.

Leverage Power of Machine Learning with ONNX

Have you ever wanted to make your apps “smarter”? This session will cover what every ML/AI developer should know about Open Neural Network Exchange (ONNX) . Why it’s important and how it can reduce friction in incorporating machine learning models to your apps. We will show how to train models using the framework of your choice, save or convert models into ONNX, and deploy to cloud and edge using a high-performance runtime.

Shaken, Not Stirred: Prompt Engineering & Semantic Kernel Mixology

Prompts play a crucial role in communicating and directing the behavior of Large Language Models (LLMs). It's like giving the model a question or topic to chat about, and it does its best to provide a relevant answer or information. Semantic Kernel is like a mixologist who looks at available ingredients and crafts new cocktail recipes. Semantic Kernel looks at different plugins, AI Prompts, and memory stores to create an execution plan. Attend this presentation to learn how Semantic Kernel can be used to build a copilot that’s creative enough to suggest new cocktails. Cheers!

Prompts play a crucial role in communicating and directing the behavior of Large Language Models (LLMs). Semantic Kernel (SK) looks at different plugins, prompts, and memory stores to create a plan. Attend this presentation to learn SK in .NET and build an agent to suggest new cocktails. Cheers!

Build Intelligent applications with ML.NET and Windows Machine Learning

Need to integrate trained machine learning models into Windows apps? In this session, we will look into how Windows Machine Learning can transform applications with the power of artificial intelligence and run it on the devices by taking full advantage of hardware acceleration. We will walk thru and learn how to train a model with ML.NET Model Builder and inference with Windows Machine Learning.

AI Agents on Demand

What do Deadpool, Ethan Hunt (Mission: Impossible) & Sherlock Holmes have in common? They’re on-demand agents — hired for a mission, bringing their skills, tools, and vanishing after the job’s done. AI Agents work the same way: purpose-built, on-demand, and mission-ready. Your mission, should you choose to accept, is to learn how to build scalable AI agents with Azure AI Agent Service in JavaScript — combining reasoning, tools & automation for next-gen cloud solutions. This message will self-destruct in 5 seconds.

Beyond Vibe Checks: Evals for Reliable AI (and Agents)

AI that can act (call tools, run code, hit APIs) is powerful—and risky. Vibe-checking outputs doesn’t scale when agents plan, execute, and adapt on their own. In this talk, we replace guesswork with evaluations: the crash-tests for agent behavior. You’ll learn a practical toolkit of code-based checks (rules and invariants), human reviews (gold-standard sampling), and model-graded evals (scalable judges with guardrails). We’ll run a live demo: define failure modes, write semantic unit tests, iterate prompts/policies, and wire the evals into CI so regressions get caught before users do. You’ll leave with a small template you can drop into your agent stacks to measure reliability, reduce cost loops, and keep actions safe—moving from “it seems fine” to “we have proof.”

From Vibes to Proof: Crash-Testing Tool-Using AI in TypeScript

AI that can act—call tools, run code, hit APIs—is powerful…and risky. Vibe-checking outputs doesn’t scale once agents plan, execute, and adapt. In this talk, we replace guesswork with evaluations: the crash-tests for agent behavior. You’ll get a practical toolkit you can ship this week: code-based checks (rules and invariants in TypeScript), human reviews (gold-standard samples), and model-graded evals (scalable judges with guardrails).
Live demo: we define failure modes, write semantic unit tests around a tool-calling agent, iterate prompts/policies, and wire the evals into CI so regressions break the build—not production. We’ll show how to capture metrics (accuracy, safety violations, cost/latency), set thresholds, and visualize drift.
You’ll leave with a small TS template you can drop into your stack to measure reliability, cut cost loops, and keep actions safe—moving from “it seems fine” to “we have proof.”