Design, build, evaluate, and govern autonomous AI agents.
© Elephant Scale
Demo connect through repositories.
A five-day, hands-on course for developers, data scientists, team leads, and project managers. You learn how AI agents plan, reason, reflect, and act — starting from a hand-built agent loop and the ReAct pattern, then adding tools, reflection, evaluation, and multi-agent coordination. Across the week you build agents on the major frameworks in production use today — LangGraph, low-code (Zapier + Custom GPTs), AWS Bedrock, DSPy + Databricks, Anthropic's Claude, and Google's ADK for voice — and you close with governance and a team capstone. Roughly half the time is lecture, half is labs, in a cloud-based, zero-install environment.
Developers, data scientists, team leads, and project managers who want to build and ship autonomous agents. Basic Python and comfort with API calls and JSON are assumed.
5 Days / ~40 hours. Intermediate. Hands-on, lab-driven (~50% labs).
- Basic understanding of machine-learning concepts.
- Familiarity with Python (preferred).
- Knowledge of API calls and JSON.
| Path | Purpose |
|---|---|
outline.md |
Full course description and module outline |
slides/ |
Markdown slide decks, one per module |
labs/ |
Lab guides, starter code, and setup docs |
course-materials/ |
Reusable templates, rubrics, checklists |
scripts/ |
Course validation helpers |
docs/ |
Supporting source documents |
images/ |
Slide images |
| Day | Module | Topic | Lab Folder(s) |
|---|---|---|---|
| 1 | 01 | Foundations of Agentic AI | labs/01-Agent-Loop-Setup, labs/02-Research-Assistant |
| 2 | 02 | Reflection & Self-Evaluation | labs/03-Reflective-Summarization |
| 2 | 03 | Tools, Structured Output & MCP | labs/04-Python-Functions-to-Tools, labs/05-Email-Assistant |
| 2 | 04 | RAG & Vector Databases | labs/06-RAG-Vector-DB |
| 3 | 05 | Evaluation, Error Analysis & Debugging | labs/07-Evaluate-Benchmark |
| 3 | 06 | Multi-Agent Systems & Planning | labs/08-Onboarding-Assistant |
| 4 | 07 | LangGraph & Tavily | labs/09-LangGraph-Essay-Writer |
| 4 | 08 | Low-Code Agentic Design | labs/10-Zapier-CustomGPT, labs/11-NoCode-CRM-Agent |
| 4 | 09 | Agentic on AWS Bedrock | labs/12-Bedrock-CRM-Assistant |
| 4 | 10 | DSPy & Databricks | labs/13-DSPy-SelfImproving |
| 4 | 11 | Agentic with Claude | labs/14-Claude-Coding-Assistant |
| 5 | 12 | Voice Agents with Google ADK | labs/15-Voice-Support-Agent |
| 5 | 13 | Governance with Databricks | labs/16-HR-Governance-Agent |
| 5 | 14 | Capstone: Enterprise AI Agent Challenge | labs/17-Capstone |
- Cloud-based, zero-install: a browser and the class VM.
- Python 3.11+ with the per-lab
requirements.txt. - API keys provided in class (OpenAI, Anthropic, Tavily) — see
labs/SETUP.md. - Cloud-console labs (AWS Bedrock, Databricks, Zapier, Google ADK) use instructor-provided accounts; each has a local-only fallback so no lab is blocked if a cloud account is unavailable.
./scripts/validate-course.sh
./labs/test-all-labs.sh
./labs/verify-setup.sh