Ship Production Agents
Build production-grade agent systems with RAG, multi-agent orchestration, and evaluation. Learn real patterns from deployed systems.
Cohort 1: Starts July 27•Cohort 2: Starts August 27•Limited to 12 participants
Course Philosophy
End-to-end production agents.
Not theory.
Real, production-grade agent systems that solve actual business problems — not theoretical projects disconnected from real-world infrastructure.
This course teaches you to build real, production-grade agent systems that solve actual business problems — not theoretical projects disconnected from real-world infrastructure.
We focus on Google Cloud Platform as our foundation because companies deploy real workloads there. You'll learn how to architect systems using GCP's native services (Vertex AI, Pub/Sub, Cloud Storage, etc.) to build scalable, observable, production-ready agent solutions. The architectural principles transfer directly to AWS, Azure, and other cloud environments once you understand the patterns.
How We Approach Agent Development
We start by building agents from scratch. You'll understand the core loop, the control flow, what actually matters. Then we move to production frameworks. LangGraph, CrewAI, Autogen — they all offer similar baseline features: model agnosticism, state management, evaluation. Those are table stakes now.
For this course, we use Google Cloud's Agentic AI Engine. Why? First-party integrations. Cloud Storage, Cloud Trace, Secret Manager, Pub/Sub — they're native, not adapters. Deployment is first-class: one command deploys to production. Many frameworks treat deployment as an afterthought. This one shipped with it.
The patterns you learn apply everywhere. Multi-cloud, AWS, Azure, or private infrastructure. The principles are the same. You'll understand agent architecture deeply enough to apply it to any platform. That's the real skill: knowing your constraints, understanding your path, and choosing the right tools for the journey ahead.
What You Build
Real agent systems: ReAct patterns, multi-agent orchestration, agent-to-agent communication (A2A), supervisor architectures. You'll understand orchestration at scale — not just toy examples.
Systems that actually ship to production.
On Real Infrastructure
Google Cloud native: Vertex AI, Pub/Sub, Cloud Storage, vector indexing, observability. External services (Qdrant, Ragas) for evaluation and distributed workflows show you how to integrate beyond a single cloud.
No vendor lock-in; principles apply everywhere.
What You Get
- Hands-on from day 1 — no lecture-only sessions
- Real agent systems architecture and patterns
- Production deployment and observability
- Capstone project you can ship
Not Covered Here
This course assumes you have foundational knowledge. We don't teach programming, data science, or statistics.
Need those fundamentals? We offer Python, ML, and advanced analytics courses on demand as external offerings.
Who is this for?
Prerequisites
- Python — solid language knowledge
- Cloud — hands-on deployment experience
- Backend or ML engineering background
- System architecture understanding
Perfect Fit
- Backend / infrastructure engineers
- ML engineers shipping to production
- Platform architects & DevOps
- Engineering leaders owning systems
6 Weeks
Curriculum
LLM Workflows to Agentic RAG
Foundational agent concepts. RAG architecture patterns. GCP deployment basics. PDF processing pipelines. Embeddings and retrieval.
Agent Loop & Orchestration
ReAct pattern implementation. Plan-and-execute workflows. Routing strategies. MCP (Model Context Protocol) standards. Tool binding.
Multi-Agent Systems
Supervisor architectures. Agent-to-agent communication (A2A). Collaborative decision-making. Shared state management. Team coordination.
RAG Evaluation & Metrics
Ragas evaluation framework. Golden test sets. Trajectory scoring. LLM-as-judge evaluation. Quality metrics for production systems.
Production Infrastructure on GCP
GCP deployment patterns. Pub/Sub messaging. Cloud Storage integration. Vertex AI backend. OpenTelemetry observability. IAM & security.
Agent Evaluation & CI/CD
Evaluation gates in deployment pipelines. Automated testing and gating. Performance thresholds. Production monitoring. Agent performance tracking.
Curriculum & Instructors
Taught by practitioners from the production frontlines
This program is designed and led by people who ship production systems at scale. Our instructors come from the front lines: startups, scale-ups, Fortune 500 enterprises. The curriculum is built from battle-tested patterns, architectural decisions, and what actually works when systems meet real production constraints.
Program led by Tomer Porat (MSc, 20+ years production systems) — bringing hands-on expertise in agentic AI workflows, system architecture, and production deployment at scale.
Every lecture, every architectural decision, every pattern you learn comes from what actually works in production. Not theory. Not tutorials. Real systems, real constraints, real solutions.
Capstone Project
Build a real production system
End-to-End Agent System on GCP
You'll build and deploy a real, production-grade agent system that handles end-to-end workflows: document processing, multi-agent orchestration, retrieval, and decision-making. This isn't a toy project — it's a system you could deploy to production with real users.
You'll learn how to:
- Architect agents with evaluation gates and observability from day one
- Orchestrate multiple agents with clear communication patterns
- Deploy to GCP with zero-downtime updates
- Evaluate agent quality with golden test sets and trajectory scoring
- Integrate external services (vector DBs, LLMs) in a distributed environment
Built on Real Cloud Infrastructure:
Google Cloud · Vertex AI · Pub/Sub · Cloud Storage · Qdrant · Ragas · MCP · OpenTelemetry · Cloud Trace
These patterns apply to AWS, Azure, and any cloud — once you understand the principles.
Proven Results
Real outcomes from real systems
Cost reduction in ML ops
Fortune 500 financial services firm — streamlined infrastructure post-consulting
Agent system to production
Startup AI infrastructure — from architecture to deployed multi-agent platform
Improvement in query throughput
Enterprise search platform — optimized vector retrieval & RAG patterns
What Leaders Say
Trusted by CXOs and engineers
VP Engineering
Series B AI startup
"Tomer helped us move from prototype to production faster than we thought possible. Not just code review — actual architecture guidance that saved us months of rework. The agent patterns we shipped are still the foundation of our platform."
→ Promoted to strategic advisor role
ML Engineer
Fintech scale-up
"The 6-week course showed me patterns I'd never seen in tutorials or bootcamps. We went from single-agent scripts to orchestrating 5 agents with observability from day one. Now I lead our platform's AI systems. Worth every dollar."
→ Promoted to Senior ML Architect 4 months post-course
Two distinct paths, one outcome: Whether you're building strategy (C-level) or shipping systems (engineering), AgenticShip accelerates your production AI journey.
Why this course is different
Exclusive Cohorts
Limited to 12 participants. Personalized feedback, direct interaction with instructors, and real collaboration. Not one of hundreds in a video course.
Your Language
Choose your cohort language: Hebrew or English. Live lectures, Q&A sessions, and instructor feedback in your preferred language. Course materials remain in English for industry-standard reference.
Cloud Patterns, Not Libraries
Learn structured cloud architecture and design patterns that apply everywhere. Specific libraries change; cloud principles are timeless. Transfer these patterns to any platform.
Built from Consulting Work
AgenticShip advises Fortune 500 enterprises and tech companies on building production AI systems. Every pattern you learn is battle-tested with real clients solving real problems.
Pricing & Admission
Reserve your seat
Ship Production Agents
6-week cohort · Limited to 12 participants
Cohort 1: Starts July 27 | Cohort 2: Starts August 27
- 6 weeks of live cohort learning
- All production code & materials
- Weekly group Q&A sessions
- Dedicated instructor feedback
- Private cohort community
- Lifetime access to materials
Early admission closes on a rolling basis. Limited to 12 participants per cohort.