← Back to Schedule

Systematic Development of Production Agents: An Evidence-Driven Approach

Saturday, April 18, 2026 · 9:00 AM – 10:30 AM · Classroom 305

Most developers start building AI agents by picking a framework, designing a multi-agent architecture, and selecting tools. This is backwards. Without evidence from real users and systematic evaluation, agents fail predictably: they solve the wrong problem, ship with unforeseen failure modes, or implement the wrong architecture.

This tutorial presents an evidence-driven framework for building agents that actually work. You’ll learn to test assumptions with prompts and users before writing production code, understand context flow before designing agent boundaries, and let evidence reveal architecture rather than imposing architecture on uncertain requirements.

The core principles we will cover:

  1. Quality risk as your north star: test your assumptions before building
  2. Evidence before architecture: playground prompts beat architecture diagrams, real user conversations beat personas, evaluation datasets beat vibe checks
  3. Right-sized decomposition: agent boundaries emerge from context clustering, not arbitrary decisions

Part 1: The Framework (30 min)

  • The Acquire-Shape-Deliver pattern for structured context engineering
  • Schema-based file system memory for maintaining coherence over long-horizon projects
  • How to identify quality risks worth testing

Part 2: Hands-on Ideation Workshop (40 min)

  • Facilitated group ideation workflow on a real problem
  • Generate structured artifacts: problem definitions, testing prompts, evaluation criteria
  • Experience systematic planning with AI development assistants (e.g., Claude Code)

Part 3: Production Case Study (20 min)

  • Walk through an agentic meal planning system built with this methodology in a few weeks
  • See the evidence trail: initial prompt failures → evaluation dataset design → context flow analysis → architecture emergence
  • Understand how structured workflows and agentic engineering principles maintained coherence across weeks of development

What You’ll Take Home:

  • Reusable ideation workflow template ready for immediate use
  • Setup guide for AI development assistants (Claude Code, MCP servers, Skills)
  • Framework for evidence-based architectural decisions

About the Speaker

Amir Feizpour

Amir Feizpour

Founder / CEO @ Aggregate Intellect Inc.

Dr. Amir Feizpour is the founder, CEO, and Chief Scientist at Aggregate Intellect building a generative business brain for service and science based companies. Amir has built and grown a global community of 5000+ AI practitioners and researchers gathered around topics in AI research, engineering, product development, and responsibility. Prior to this, Amir was an NLP Product Lead at Royal Bank of Canada. Amir held a research position at University of Oxford conducting experiments on quantum computing resulting in high profile publications and patents. Amir holds a PhD in Physics from University of Toronto. Amir also serves the AI ecosystem as an advisor at MaRS Discovery District and entrepreneur in residence at North Forge and Lab2MArket, and works with several startups as fractional chief AI officer. Amir engages with a wide range of community audiences (business executives to hands-on developers) through training and educational programs as well. Amir leads Aggregate Intellect’s R&D via several academic collaborations with McGill U., University of Toronto, and Alberta Machine Intelligence Institute.