AR Academy · Cohort 1
Shipping Production Agentic Workflows
A 6-week cohort for senior engineers and tech leads. Taught by Alvin Reyes (AR Data — HP, Scotiabank, Macquarie, Protocol Labs).
The problem
Most agent systems work in the demo and break in production.
The failures are predictable: brittle prompts, eval frameworks that don't catch real failure modes, cost spikes nobody saw coming, non-deterministic bugs that surface at 3am, and architecture choices that lock you into a single model vendor.
I've spent the last year focused on shipping production agentic systems, drawing on 20 years of enterprise engineering across Oracle, IBM, HP, Protocol Labs, Macquarie, and Scotiabank. This cohort is the playbook.
Who this is for
If you see yourself here, you'll get full value from every session.
Senior engineers (3+ years)
Building or about to build agent systems your company will actually depend on.
Tech leads
Who need their team to ship agentic workflows without burning out on broken demos.
Founders / CTOs at AI-native startups
Who want production-grade habits in the team before scaling — not after the first incident.
Who this is not for
No hard feelings. Just want to be clear.
- Total beginners — this assumes you can read code and have shipped something.
- People looking for "AI strategy" framing without code.
- Anyone expecting a self-paced, watch-and-go course. This is live, interactive, and you will ship.
What you'll learn
Drawing on a year of focused agentic work and 20 years of broader production engineering patterns applied to agents. Code-heavy. Theory only where it serves the build.
The Production Mental Model
Why agents fail, what production really means for non-deterministic systems, mapping the surface area. Drawing on a year of focused agentic work and 20 years of broader production engineering patterns applied to agents.
Architecture Patterns That Hold Up
Tool use, orchestration, state management, retries, fallbacks, deterministic boundaries. Patterns that don't fall over at week 3.
Eval Frameworks That Catch Real Failures
Beyond vibes-based testing. Building eval suites that catch regression before deploy. What the field has figured out so far, what it hasn't, and how to think about evaluation honestly.
Observability and Cost Control
Tracing, logging, replay, cost budgets. The metrics that actually predict cost spikes. Practical patterns from running agentic systems in production.
Deploy, Rollback, and Non-Determinism
Versioning prompts, models, and tools. Canary deploys for agents. Rollback strategies when "rollback the model" isn't an option.
Capstone Reviews
Each student presents the agent system they built. Live feedback from Alvin and the cohort. Ship something you'd put your name on.
Format
Instructor
Alvin Reyes. The AR in AR Data.
20 years of enterprise engineering across Oracle, IBM, HP, Protocol Labs, Macquarie, and Scotiabank. Founded AR Data; bootstrapped, shipping production AI and blockchain systems for HP, Scotiabank, Macquarie, and Protocol Labs.
I teach what I actually do, not what looks good on a slide.
Instructor pedigree
20 years of enterprise engineering across these teams informs the patterns in this course.
Pedigree logo row placeholder — Oracle, IBM, HP, Protocol Labs, Macquarie, Scotiabank
AR Data clients
AR Data has shipped production systems for these organizations.
Our engineers have shipped at
Our team has delivered at some of the world's most demanding organizations — enterprise banks, global infrastructure firms, and leading decentralized infrastructure projects.
FAQ
Reserve your seat
20 seats. First cohort pricing available now.
Questions? alvin@ardata.tech or DM on LinkedIn.
