LumenAIStart learning
Back to home
Live bootcamp

Build Agentic Apps Bootcamp

A hands-on 2-day sprint that takes you from LLM basics to a deployed AI agent — with tools, memory, RAG, guardrails, and evals baked in.

Duration
2 days
Daily hours
9:00 – 18:00
Format
Live + hands-on labs

Download syllabus & resources

Grab the full syllabus PDF and a resources bundle with starter repos, prompt templates, an eval harness, and a curated reading list.

Syllabus (PDF)
2-day schedule, outcomes, pre-work
Resources bundle (ZIP)
Starter repos, prompts, evals, reading list

What you'll build

By the end of Day 2 you'll have shipped a working agent that plans, calls tools, retrieves grounded knowledge, remembers users across sessions, and runs behind a real API with logging and evaluation in place.

Learning outcomes

  • Design agent loops with planning, tool use, and memory
  • Implement function calling and structured outputs reliably
  • Build a RAG pipeline with citations and freshness controls
  • Add guardrails against prompt injection and unsafe outputs
  • Write evals that catch regressions before users do
  • Deploy an agent behind a real API with observability

Day 1 — Foundations & your first agent

  1. Agentic AI, demystified

    09:00 – 10:00

    What separates an 'agent' from a chatbot: planning, tool use, memory, and autonomy loops.

  2. LLM foundations for builders

    10:00 – 11:30

    Tokens, context windows, structured outputs, function calling, and cost/latency trade-offs.

  3. Prompt engineering that scales

    11:30 – 13:00

    System prompts, few-shot patterns, chain-of-thought, and prompt versioning.

  4. Tools & function calling (lab)

    14:00 – 15:30

    Wire your first agent to real tools — web search, calculators, and a custom API.

  5. RAG for grounded answers (lab)

    15:30 – 17:00

    Chunking, embeddings, vector search, and retrieval-augmented generation with citations.

  6. Day 1 wrap + Q&A

    17:00 – 18:00

    Review builds, debug common failure modes, and preview Day 2.

Day 2 — Memory, safety & shipping

  1. Agent architectures

    09:00 – 10:30

    ReAct, plan-and-execute, multi-agent, and when to use each pattern.

  2. Memory & state (lab)

    10:30 – 12:00

    Short-term context, long-term vector memory, and per-user personalization.

  3. Guardrails & safety

    12:00 – 13:00

    Prompt injection defenses, output validation, PII handling, and human-in-the-loop.

  4. Evaluation & observability

    14:00 – 15:30

    Build an eval suite, trace runs, track regressions, and score with LLM-as-judge.

  5. Ship it (capstone)

    15:30 – 17:00

    Deploy your agent behind an API + minimal UI, with auth, rate limits, and logging.

  6. Demos & next steps

    17:00 – 18:00

    Everyone demos, peer feedback, and a roadmap for taking your agent to production.

Who this is for

  • Software engineers adding AI features to real products
  • Data/ML practitioners moving from notebooks to shipped apps
  • Technical founders prototyping agentic products
  • Curious builders comfortable with JavaScript or Python

What you'll need

  • • A laptop with Node.js 20+ or Python 3.11+ installed
  • • A code editor (VS Code recommended)
  • • An OpenAI, Anthropic, or Lovable AI Gateway API key (we'll help you set one up)
  • • Basic familiarity with JavaScript or Python — no ML background required

Tech stack you'll touch

LLM APIsFunction callingEmbeddingsVector DBRAGLangChain / LlamaIndexEvalsObservabilityREST APIs

Ready to build your first agent?

Subscribe to be notified about the next cohort date and get the pre-work reading list.

Explore the curriculum