Just a preview. More workshops to be announced!
With Marina Wyss
AI systems are notoriously hard to evaluate: They’re non-deterministic, quality is often subjective, and a pipeline with multiple steps like retrieval, prompts, and tool calls can fail silently. After coaching hundreds of engineers into AI, the skill I see missing most often is structured evaluation. Teams that are newer to AI tend to rely on spot-checks and vibes, and don’t know there’s a problem until a customer complains or something embarrassing ends up on Twitter. This workshop aims to fix that. We’ll cover evaluation from absolute beginner to production: picking metrics, building your first eval suite, component-level evals for everything from RAG to tool use, using LLMs as judges effectively, and evaluating multi-agent systems in production. You’ll write real evals during the session and leave with systems you can reuse at work and a way to feel confident in what you’re building.
Michael Hladky
88% of organizations now use AI in at least one business function, yet most remain stuck in the pilot phase — and over half have already experienced negative consequences from AI inaccuracy. Without structured adoption, teams face inconsistent output, hidden security risks, and costly refactoring cycles. The tool is not the problem; the approach is. Understanding the mechanics behind AI code generation is the difference between frustration and flow!
In this workshop, you will uncover the principles that power every AI coding assistant. You will learn why context gets lost mid-conversation, why hallucinations happen, and how to structure your interactions to get reliable and secure results. These fundamentals apply regardless of which tool sits in your editor.
Beyond single prompts, you will explore the different ways to work with AI: agent workflows and spec-driven development as the daily default, chat completions and inline suggestions where they still shine, and an outlook on running multiple agents in parallel. Understanding these modes — and when to reach for each — is the difference between fighting your tools and flowing with them.
Because modern assistants generate code faster than anyone can read it, review — not generation — becomes the bottleneck. You will learn to make verification a first-class part of your workflow: critique runs, quality gates, and checks that keep pace with what your agents produce.
You will build a complete AI context ecosystem for your project: architecture documentation, style guides, context files, instruction configurations, and MCP integrations that connect your AI to external tools and data sources. Following open standards like Agent Skills, this infrastructure stays portable across tools – it doesn’t just help you today, it empowers those who come after. The foundations you create improve any assistant’s understanding of your codebase – and position you to adapt as tools evolve.
This is a hands-on workshop. You will work through exercises in your own project, building the infrastructure that makes AI-assisted development reliable. No theoretical lectures – you leave with real files you can use immediately.
We break each concept into manageable pieces with exercises that reinforce what you learn.
Prerequisites
– Basic development experience in Angular (examples use Angular, but AI concepts apply to any stack)
– Familiarity with your preferred code editor
– Access to any AI coding assistant (Copilot, Claude Code, Cursor, Codex, Kiro, or similar)
– No prior AI or machine learning knowledge required
Level: Intermediate