ESDS Advanced follows the same rhythm as the foundational ESDS course:
lecture and practice in the mornings, project time in the afternoons.
Daily structure
- Mornings: instructor-led lectures interleaved with hands-on practice on
the day's topic.
- Afternoons: team project time, applying the morning's skills to a real
NOAA/NCICS-related problem.
Weekly structure
The course spans three weeks, one module per week:
- Week 1 — Agentic Cowork Tools. The first two days cover how LLMs and
agents work; the next three focus on Claude Code, scaling to parallel
instances with remote MCP.
- Week 2 — Agentic Systems. Building a REACT loop and a simple harness,
building good skills, then building a full agentic workflow system.
- Week 3 — Deep Learning. Building and training a PyTorch model on AWS
GPUs, then using and improving open-source models and evaluating results.
Presentations
Projects are presented weekly, with a final presentation at the end of
the course. The goal is to find an aspect in each of the three modules and put
them together into a single final product.
Getting help
As in the foundational course, we encourage a hierarchy of resources when you
hit a problem:
- Your programming and cowork assistants.
- Search and documentation.
- Your team.
- Course instructors.
Learning to get unblocked efficiently — including knowing when to escalate from
an agent to a human — is itself one of the skills this course teaches.