LLM Agents & Multi-Agent Swarms

LangGraph, state graphs, reflection loops, multi-agent delegation, and human-in-the-loop.

3 weeks · wk 15–17Applied pace5 topics15 Interview Questions

Why this pace:Autonomous agents move beyond single-turn answers into multi-step goal completion. Building resilient agents requires cyclic state machines, tool orchestration, persistent memory, and deterministic guardrails.

SEC FOCUS

Autonomous agents with write/execute permissions (e.g. bash execution, SQL queries, payments) must have strict permission boundaries, rate limits, and mandatory human confirmation gates.

Topics & Action Items (5)

Agent Cognitive Architectures: ReAct, Plan-and-Solve & Reflexion

5 hrs

Stateful Agent Graphs with LangGraph

6 hrs

Stateful Memory: Short-Term, Long-Term & Ephemeral (Mem0, Zep)

4 hrs

Human-in-the-Loop (HITL) & Resumable Async Workflows

4 hrs

Multi-Agent Collaboration: CrewAI, AutoGen & Swarm Consensus

5 hrs

Previous Stage

Advanced Retrieval-Augmented Generation (RAG)

Next Stage

Model Fine-Tuning & Parameter-Efficient Adaptation