Back to hands-on classes
Coming upAdvancedOnline

AI-Powered Software Architect: Designing Scalable Agentic Systems

Move beyond basic prompting. Learn to design, architect, and deploy complex, reliable multi-agent systems at scale using modern orchestration frameworks.

Schedule

August 29 - September 26, 2026

Duration

4 Weeks • 4hrs/wk

Project

Hands-on capstone

Detailed Curriculum

4 practical sections built around live exercises.

01

Architecture for agentic workflows

Design agent systems with explicit boundaries, state, and responsibilities.

Topics covered

  • Workflow vs autonomous agent decisions
  • State machines and graph patterns
  • Service boundaries
  • Failure and fallback design

Hands-on lab

Create an architecture diagram and responsibility map for a production agent workflow.

02

Multi-agent orchestration

Coordinate specialized agents without losing control of cost, latency, or reliability.

Topics covered

  • Supervisor-worker patterns
  • Planner-executor systems
  • Task delegation
  • Shared memory and coordination risks

Hands-on lab

Prototype a multi-agent workflow with explicit routing and validation steps.

03

MCP and production integrations

Connect agents to data and tools through contracts that can be governed.

Topics covered

  • MCP server boundaries
  • Tool schemas
  • Authorization and audit trails
  • Integration testing

Hands-on lab

Design an MCP-backed tool layer for one enterprise workflow.

04

Evaluation, observability, and deployment

Ship agent systems with the instrumentation needed to improve them safely.

Topics covered

  • LangSmith tracing
  • RAGAS evaluation
  • Golden datasets
  • Deployment and rollback strategy

Hands-on lab

Define an evaluation plan, trace map, and deployment checklist for the capstone architecture.

What You Get Out Of It

Concrete capabilities you should leave with.

Architect scalable agentic systems

Choose orchestration patterns for real constraints

Plan MCP-enabled tool integration safely

Design observability, evaluation, and deployment workflows