Frontier Agentic Engineering Certificate for Engineers & Technical PMs
Henry Shi
Member of Technical Staff @ Anthropic
Shubham Saboo
Sr. AI PM @ Google (#1 AI Agent GitHub)
Every other course teaches you to build agents. This is where building ends.
"Vibe coding raised the floor. Frontier agent engineering protects the ceiling." — Andrej Karpathy
For 10 years, Product Faculty has trained 100,000+ product and engineering leaders. Cohort-based programs, capstone mandatory, designed to finish what they start.
This is our flagship for senior engineers and technical PMs whose agents are already in production. Six weeks. Taught and capstone-reviewed by Henry Shi (Anthropic) and Shubham Saboo (Google Cloud, creator of Awesome LLM Apps — #1 agent repo on GitHub).
Most agent courses teach you how to build. This one starts where building ends. You walk in with a real problem. You walk out with a 24/7 agent fleet solving it while you sleep.
Team (10+): Private Cohort Portal for internal IP + 1:1 support. support@productfaculty.com.
⚡ BEST VALUE — AI Builder Track: Product Faculty's top AI courses in one track. 24-month all-access, retake any cohort (over $10K value).
• AI Product Management ($2,500)
• AI Product Strategy for Leaders ($2,500)
• AI Product Leadership ($5,000)
• Frontier Agentic Engineering ($2,500) — this course.
👉 AI Builder Track $3,995: enroll in this course first → upgrade details in your welcome email.
What you’ll learn
Go from agent builder to agent manager. Run a fleet that ships work while you sleep, with the discipline to prove it in production.
Brief sub-agents with the same instinct a senior leader brings to onboarding a new hire
Write briefs precise enough that agents execute without hand-holding
Apply the new-hire onboarding model: some context, some examples, clear rules of engagement
Design memory tiers (working, episodic, shared) so the agent doesn't start dumb every session
Author custom Skills as the layer where context becomes capability
Move context across Codex, Claude Code, OpenClaw, and Hermes without losing capability
Review agent work structurally, not line by line, like a manager reviews a senior IC
Run a two-agent pattern: one writes, one reviews, with explicit acceptance criteria
Use corrective prompt engineering to turn every correction into a permanent rule
Offline evals for release confidence; online monitors for production drift
User simulators that push back, contradict, and pressure-test your agent
Failure clustering and LLM-as-judge councils that turn scores into product fixes
Design the org chart for the agent team the way you'd design it for humans
Choose between single-agent-with-tools and multi-agent based on production tradeoffs
Resolve conflicts when agents disagree, with documented escalation paths
Heartbeat self-healing, cost telemetry, scoped credentials, and prompt injection defense
Design the handoff between human and fleet: what queues, what wakes you, what self-resolves
Ship a capstone fleet with a 72-hour unattended run and an incident log to prove it
Learn directly from Henry & Shubham
Henry Shi
Member of Technical Staff @ Anthropic. Runs production agent fleets at scale.
Shubham Saboo
Sr. AI PM @ Google | Awesome LLM Apps (#1 GitHub agents repo, 109k⭐)
Who this course is for
Senior software or ML engineers (3–7 years) at AI-native startups who shipped an LLM feature and hit the reliability wall.
Staff engineers and technical PMs making architectural decisions on AI systems without writing all the code themselves.
Engineering leaders who want to be the go-to AI architect on their team and run a fleet of agents the way they run a team of humans.
Prerequisites
Shipped one LLM feature to production
You should already know what it feels like to hit the reliability wall. This course starts where building ends.
3+ years as a senior software or ML engineer
We assume the engineering instincts that come with seniority — system design, code review, and shipping under constraints.
Foundational agent vocabulary (e.g. DeepLearning.AI Agentic AI)
If you can't define context windows, tool use, and agent loops, do a foundational course first. Otherwise you'll be lost by Week 2.
What's included
Live sessions
Learn directly from Henry Shi & Shubham Saboo in a real-time, interactive format.
Lifetime access
Go back to course content and recordings whenever you need to.
Community of peers
Stay accountable and share insights with like-minded professionals.
Certificate of completion
Share your new skills with your employer or on LinkedIn.
Maven Guarantee
Your purchase is backed by the Maven Guarantee.
Course syllabus
Schedule
Live sessions
2 hrs / week
Live weekly session with the lead instructor for that week (Henry or Shubham), plus alternating office hours.
Projects
4 hrs / week
Weekly capstone build, code reviews in cohort, peer rubric. Each week adds one layer to your fleet.
Async content
2 hrs / week
Module material between sessions: readings, walkthroughs of context, skills, evals, and multi-agent patterns.