Day 1 — Monday, July 27, 2026
Agentic coworking and the anatomy of an agentic system
Week 2 Introduction
30 min- Course Recap
- Week 2 Introduction
- Week Goals & Objectives
Advanced Agentic Coworking
50 min- Parallelizing your work across multiple agents
- Git worktrees: isolated workspaces for concurrent agents
- Managing, monitoring, and merging parallel agent sessions
Introduction to Agentic Systems
50 min- From using agents to building them
- The anatomy of an agentic system
- Where agentic systems pay off in scientific work (and where they don't)
Logistics & API Key Setup
25 min- Accounts, access, and environment check
- Getting and configuring your API keys
- Verifying your setup before Day 2
Team Project Time & Office Hours
60 min- Team project work time
- Office hours — get your questions answered
- Mentoring and hands-on practice
Day 2 — Tuesday, July 28, 2026
Building agentic workflows
Frameworks: LangChain, Bedrock, and Langfuse
50 min- Bedrock as the model endpoint, LangChain as the uniform interface
- Your first call to Claude Haiku, and what it costs
- Adding tracing, and reading your first trace
Structured Output, Tools, and LangGraph
50 min- Typed results instead of text to parse
- Tool calling: letting the model call your Python functions
- LangGraph: state that persists, and a loop that terminates
Capstone: Finding Relevant Research
50 min- Deterministic workflows versus autonomous agents
- Hands-on: a publication finder for the NCICS catalogue
- Bonus: expose it as a local MCP server for Cowork
Team Project Time & Office Hours
60 min- Team project work time
- Office hours — get your questions answered
- Mentoring and hands-on practice
Day 3 — Wednesday, July 29, 2026
RAG and advanced agentic workflows
Day 2 Recap
30 min- Where we got to: LangChain to Bedrock, tracing, typed output
- Questions from yesterday's exercises
- What is still ahead: tools, memory, loops, MCP
Retrieval-Augmented Generation
50 min- What RAG is, and when retrieval beats a longer prompt
- Chunking, embeddings, and vector search
- Connecting your agents to real data
- RAG inside an agentic system: retrieval as a tool
Advanced Agentic Workflows: Memory, Loops, and MCP
80 min- Tool calling: letting the model call your Python functions
- LangGraph: state that persists, and loops that terminate
- Types of memory, storage options, and context management
- Building your own MCP server
Team Project Time & Office Hours
60 min- Team project work time
- Office hours — get your questions answered
- Mentoring and hands-on practice
Day 4 — Thursday, July 30, 2026
Memory, loops, and observability
Day 3 Recap
25 min- Where we got to: RAG, and retrieval as a tool
- Questions from yesterday's exercises
- What is still ahead: memory, loops, and reading a trace
Memory and Looping in Agentic Workflows
80 min- Types of memory, storage options, and context management
- LangGraph: state that persists across steps
- Loops that terminate, and recursion limits
- Putting memory and loops together in one workflow
Introduction to Langfuse
55 min- What a trace is: span, generation, and event
- Reading the Langfuse UI
- Diagnosing a slowdown from the trace tree
Team Project Time & Office Hours
60 min- Team project work time
- Office hours — get your questions answered
- Mentoring and hands-on practice
Day 5 — Friday, July 31, 2026
Observability, monitoring, and security
Day 4 Recap
30 min- Where we got to: memory, loops, and a first look at Langfuse
- Questions from yesterday's exercises
- What is still ahead: monitoring and securing what you built
Langfuse, Monitoring, and Security
100 min- Reading traces and spans in Langfuse
- Monitoring cost, latency, and failure modes
- Diagnosing a slowdown from the trace tree
- Securing agentic systems: secrets, keys, and data handling
- Human-in-the-loop and approval workflows
Week 2 Close-Out
30 min- What we built this week
- Where to go from here
- Keeping your environment running after the course