Agents & workflows
Build a multi-step agent workflow
Plan-act-observe loops, state, retries, human-in-the-loop, with LangGraph/OpenAI Agents/Claude SDK or plain code.
~20 focused hours·intermediate
Tools: LangGraph, OpenAI Agents SDK, CrewAI, Claude Agent SDK, Python asyncio
What employers mean
You should be able to…
- Implement a plan-act-observe (ReAct-style) loop where the agent reasons, calls a tool, observes the result, and decides the next step
- Persist agent state (conversation history, scratchpad, intermediate results) across multiple steps and turns
- Add retry and error-recovery logic when a step fails (bad tool output, model refusal, timeout)
- Insert a human-in-the-loop checkpoint before a risky or irreversible action
- Choose between a fixed workflow (deterministic DAG) and a fully autonomous agent loop based on task predictability
- Coordinate multiple specialized agents (e.g. researcher + writer + reviewer) with a defined handoff protocol
- Set hard step/iteration limits and budget caps so an agent loop can't run away
- Stream intermediate agent steps to a UI so users see progress, not just a final answer
Needs first: Implement tool / function calling
Learn — free, link-checked
The few resources that matter
Read · beginner · 25 min · python.langchain.com
Tool calling concepts
Explains the provider-agnostic tool-calling abstraction that LangChain/LangGraph agents are built on. — LangChain
Course · beginner · 300 min · huggingface.co
Welcome to the AI Agents Course
Free multi-unit course covering agent fundamentals, smolagents, LangGraph and multi-agent systems with a certification project. — Hugging Face
Read · intermediate · 30 min · anthropic.com
Building Effective AI Agents
Defines the workflow-vs-agent patterns (routing, orchestrator-worker, evaluator-optimizer) employers expect you to name and choose between. — Anthropic
Read · intermediate · 40 min · openai.github.io
OpenAI Agents SDK
Lightweight alternative to LangGraph for agent loops, handoffs and guardrails, increasingly named in JDs alongside LangGraph/CrewAI. — OpenAI
Read · intermediate · 60 min · langchain-ai.github.io
LangGraph Quickstart
Hands-on build of a stateful, cyclic agent graph with memory and human-in-the-loop, the most-requested framework in Indian GenAI JDs. — LangChain
Course · intermediate · 90 min · deeplearning.ai
MCP: Build Rich-Context AI Apps with Anthropic
Builds an MCP server and connects it to a chatbot client, covering resources, prompts and tools in one project. — DeepLearning.AI (with Anthropic)
Course · intermediate · 120 min · deeplearning.ai
AI Agents in LangGraph
Short video course building a ReAct agent and a LangGraph agent from scratch, taught by LangChain's founder. — DeepLearning.AI (with LangChain)
Course · intermediate · 120 min · deeplearning.ai
Multi AI Agent Systems with crewAI
Builds role-based multi-agent crews, a framework named directly in Indian GenAI JDs, for a realistic research/writing pipeline. — DeepLearning.AI
Practice
Multi-agent NBFC loan-document triage workflow
Build a LangGraph workflow that ingests a loan application (PAN card, salary slip, bank statement PDFs), routes it through an extraction agent, a policy-check agent, and a summarizer agent, and pauses for human approval before marking an application 'approved'. Persist state so the workflow can resume after the human responds. Cap the loop at a fixed number of steps and log every transition.
Done when
- A LangGraph (or equivalent) state graph with at least 3 distinct agent nodes and conditional routing
- A human-in-the-loop interrupt before the final approval decision, with the workflow resuming correctly after input
- State persists across a process restart (checkpointing to disk/DB)
- A step/iteration cap enforced so the graph cannot loop indefinitely, with a logged trace of a full run
Prove it
Evidence a recruiter can check
- Public GitHub repo with an architecture diagram of the agent graph and a README explaining each node
- A recorded trace/log of one full run showing the human-in-the-loop pause and resume
- A short write-up of one failure mode you hit and how you fixed it
Interview
Questions you'll get asked
- When would you use a fixed pipeline instead of a full autonomous agent, and why?
- How do you stop an agent loop from looping forever or blowing through its token budget?
- Design a multi-agent system for processing loan applications — what agents, what do they hand off to each other?
- How would you add a human-approval step before an agent sends an email or makes a payment?
- Explain LangGraph's state graph model — nodes, edges, and conditional routing.
- How do you recover an agent's state after a crash mid-workflow?
- What failure modes have you seen in production agents, and how did you fix them?
- How do you decide how much autonomy to give an agent vs hard-coding the next step?