← All Labs Tracks
Open Lesson →
Open Lesson →
Curriculum Track
•
~14 Hours
Agentic AI
Deterministic control loops, tool interfaces, and failure isolation for autonomous workflows
A structured engineering curriculum covering the transition from simple prompt-response wrappers to deterministic execution loops, schema validation barriers, bounded recursion, and offline trajectory evaluations.
Start Module 1: Deterministic Control Loops vs Autonomous Reasoning →
Expected Prerequisites
- • Working proficiency in TypeScript or Python
- • Basic familiarity with LLM APIs (prompting, token limits, structured output mode)
- • Fundamental understanding of state machines and API idempotency
Architectural Invariants Covered
- ✓ Design deterministic Directed Acyclic Graph (DAG) state machines for multi-step agent workflows
- ✓ Enforce strict JSON Schema validation boundaries on LLM tool invocations with additionalProperties: false
- ✓ Implement circuit breaker governors to prevent runaway recursion and token-spend depletion
- ✓ Build replayable offline evaluation harnesses to detect non-deterministic agent decision degradation
Detailed Curriculum Syllabus
1 Modules • 2 Lessons
MODULE 01
Module 1: Control Loops & State Machines
Moving beyond naive ReAct loops into bounded, deterministic execution graphs.
Lesson 01 • 20 minutes
Deterministic Control Loops vs Autonomous Reasoning Classify workflows into fixed state machines vs open-ended loops, and identify the point where unstructured autonomy becomes an operational liability.
Lesson 02 • 25 minutes
Bounding Execution Loops & Tool Calling Schemas Author unambiguous tool definitions using strict schema constraints and implement pre-flight argument validation.