Interview prep for the AI era
Prepare for AI agent interviews end to end: learn agent fundamentals and LangGraph, build a production agent, deploy it to GCP, AWS or Azure, and practise 59 interview questions with model answers and code.
What's inside
Six tracks. Do them in order or jump straight to what you need.
Agent Fundamentals
LLMs, tool calling, ReAct, planning, memory, RAG, multi-agent patterns and when not to use an agent.
LangGraph Deep Dive
State, nodes, edges, reducers, checkpointers, human-in-the-loop, streaming, subgraphs and supervisors.
Real-World Project
Build "SupportPilot": a customer-support agent with tools, RAG, Postgres memory, approvals, FastAPI, Docker and tests.
Deploy to GCP
Cloud Run, Artifact Registry, Cloud SQL, Secret Manager and GitHub Actions with Workload Identity Federation.
Deploy to AWS
ECR, ECS Fargate behind an ALB, RDS Postgres, Secrets Manager, CloudWatch and GitHub OIDC.
Deploy to Azure
Azure Container Apps, ACR, PostgreSQL Flexible Server, Key Vault, managed identity and GitHub Actions.
Agent Ops
Orchestration patterns, state & monitoring, scaling, failure recovery and guardrails, with real-world incidents and interview questions.
Interview Q&A
Searchable, filterable bank of questions with model answers, from fundamentals to system design. Includes flashcard mode.
Suggested 4-week roadmap
About 6–8 hours per week. Tick off steps as you go — progress is saved in your browser.
Week 1 — Foundations
Read Fundamentals. Write a raw tool-calling loop in ~50 lines of Python without any framework so you understand what frameworks hide.
Week 1–2 — LangGraph
Work through LangGraph: build a ReAct graph by hand, add a checkpointer, an interrupt, and stream tokens.
Week 2–3 — Build SupportPilot
Follow the project: tools, RAG, approvals, API, Docker, evals. Push it to your GitHub — it's your portfolio piece.
Week 3 — Deploy to one cloud (then a second)
Pick the cloud your target employer uses: GCP, AWS or Azure. Then skim the other two so you can compare them in an interview.
Week 3–4 — Run it like production
Work through Agent Ops: durable execution, monitoring, capacity maths, failure recovery and layered guardrails. This is what separates senior candidates.
Week 4 — Interview prep
Drill the Q&A bank in flashcard mode. Practise the system-design questions out loud, using your project as the worked example.
The project you'll build
One codebase, three clouds.