Agentic AI Bootcamp: from prompt to production agent.
Five intensive days. Your engineers build a working, evaluated, governed agent on your own data — not a toy demo. RAG, tool design, multi-agent orchestration, evals, guardrails, and a production capstone.
What this bootcamp is
Most "AI training" ends at prompt engineering. This bootcamp starts where production begins: agents that call tools, read your data, recover from failures, and get evaluated like software. Every concept is paired with a lab; the week ends with a capstone your team can actually deploy.
Who should attend
- Software, data, and ML engineers tasked with building AI features
- Architects designing agent platforms or AI product strategy
- Technical leads who need to evaluate agent frameworks and vendors
What you need coming in
- Professional Python; comfort with APIs and JSON
- Basic familiarity with LLMs (you've used an API or a chat product)
- A laptop with Docker; cloud credentials provided for labs
- No prior agent-framework experience required
Learning objectives
- Design agent tool schemas and orchestration loops that handle failure gracefully
- Build RAG pipelines with chunking, retrieval, and reranking tuned by measurement
- Implement multi-agent workflows with planning, delegation, and human-in-the-loop gates
- Write eval suites — golden datasets, LLM-as-judge, and regression gates — for agent behavior
- Apply guardrails: prompt-injection defense, PII handling, policy gates, and audit logging
- Cost and latency engineering: model routing, caching, and streaming for production budgets
Day by day
Mornings are concept + live build; afternoons are guided labs. Evenings are optional office hours.
Day 1 — Agent fundamentals: the loop that does work
How agents actually work: reasoning loops, tool calling, structured output, and memory. Framework landscape (LangGraph, CrewAI, raw SDKs) with honest tradeoffs. When not to use an agent.
LABS → build a tool-calling agent from the raw API; add structured output + retries; benchmark loop vs. single-shot on a task setDay 2 — RAG that survives contact with real documents
Chunking strategies by document type, embedding selection, hybrid retrieval (dense + BM25), reranking, and citation generation. Measuring retrieval quality instead of eyeballing it.
LABS → build a RAG pipeline over a messy corpus; tune chunking with retrieval metrics; add reranking and measure the liftDay 3 — Multi-agent orchestration
Planner–executor patterns, agent handoffs, shared state vs. message passing, and human-in-the-loop gates. Debugging distributed agent behavior with traces.
LABS → build a 3-agent research pipeline (planner, researcher, critic); add approval gates; trace and fix a failure cascadeDay 4 — Evals, guardrails, and production hardening
Golden datasets, LLM-as-judge evals, regression suites on every prompt change. Prompt-injection defense, PII redaction, policy gates, audit logging. Cost/latency engineering: routing, caching, streaming.
LABS → write an eval suite for your Day-3 agent; red-team it with injection attempts; add guardrails and re-run the suite greenDay 5 — Capstone: ship it
Teams build an end-to-end agent on a real use case (bring your own data or use ours): scoped tools, RAG over your docs, evals, guardrails, and a deployment plan. Final demos with production-readiness review from the instructors.
CAPSTONE → working agent + eval report + guardrail checklist + deployment plan; demo and reviewDelivery options
5-day bootcamp
The full syllabus above, onsite at your office or virtual. Up to 20 participants; 2 instructors. Includes lab environment, all materials, and a post-course office-hours session.
3-day intensive
Days 1–3 plus a condensed evals/guardrails module. For teams that need builders fast and will learn production hardening on the job with our follow-on support.
1-day exec workshop
For leadership: the agent landscape, build-vs-buy, governance and risk, and what "production-ready" actually costs. No coding; plenty of architecture.
Private cohorts only — we don't run public dates. Content is tailored to your stack (cloud, frameworks, and data) during a pre-course scoping call. Related: deploy an AI workforce with our engineers, or browse the Academy.
Train your team on agents that ship.
Tell us your team size, timeline, and stack — we'll scope the cohort and send a proposal.