Getting Started

  • Welcome & Overview
  • Logistics & Format
  • Team Participation
  • Prerequisites
  • Setup & Day 0 Checklist

Week 1 — Agentic Cowork Tools

  • Week 1 Overview
  • Week 1 Schedule
  • Week 1 Resources

Week 2 — Agentic Systems

  • Week 2 Overview
  • Week 2 Schedule
  • Week 2 Resources

Week 3 — Deep Learning

  • Week 3 Overview
  • Week 3 Schedule
  • Week 3 Resources

Week 4 — Foundation Models in Research

  • Week 4 Overview
  • Week 4 Schedule
  • Week 4 Resources
ESDS Advanced/Getting Started/Welcome & Overview
OverviewESDS CourseAll Programs

Up next

Logistics & Format

Dates, location (Asheville + Zoom), daily rhythm, cost and equipment, attendance, and how to get help.

On this page

    Welcome & Overview

    ESDS Advanced is a four-week applied intensive from the North Carolina Institute for Climate Studies (NCICS), running in two parts: two weeks in summer 2026 (the back-to-back weeks of July 20 and July 27) and two weeks in winter 2026–27 (dates TBD). It has one goal: to make you faster and more effective as a scientist by putting generative AI to work in your research — and giving you the judgment to trust the results.

    This is not a survey course. Every week mixes lectures with hands-on practice on real problems, and you finish with three concrete capabilities:

    1. Work faster with agentic tools. Use Claude, Claude Cowork, and Claude Code to research, write, and build scientific software faster with better quality — while validating everything they produce.
    2. Automate real work. Design, build, and launch agentic workflow systems of your own, with monitoring, security, and a human in the loop.
    3. Build and use models. Architect, build, train, and tune deep learning models — then build on foundation models, evaluate them rigorously, and incorporate them intelligibly into your research.

    Scientific Quality comes First

    The philosophy behind this course is simple: AI should raise the standard of your science, not lower it. Every tool and technique we teach is paired with validation — checking agent output against ground truth, evaluating models rigorously, and telling quality work from AI slop. Speed without trust isn't science, and we deliberately designed this course to prioritize scientific rigor over speed shortcuts.

    The structure of the course that will be repeated throughout each week is: 1) Teach the fundamentals so you understand what is happening under the hood, 2) Show you what the tools can do with hands on practice, and 3) Demonstrate all the failure modes, caveats, limitations, and danger zones you need to know to use them safely and effectively.

    The core skill we want you to leave with is Discernment the ability to discriminate between good and bad AI output and know intuitively and explicitly when and how to use AI tools effectively to achieve your scientific goals.

    Who it's for & how to register

    Register for ESDS Advanced here — please register before the course begins on July 20, 2026.

    • The course is targeted at NOAA and NCICS staff, but is not exclusive to them. If you're interested and not sure you fit, register anyway — the instructors will reach out.
    • There is no cost to NOAA and NCICS participants: compute and AI usage are covered through a partnership with AWS.
    • Partial attendance is allowed. You can take Week 1 on its own; Weeks 2 through 4 each require Week 1 first.
    • Attend in person (Veach-Baley building, Asheville) or remote over Zoom — recordings are posted to this site.
    • Team participation is optional. Team projects are a great way to get hands-on practice, but you don't need one to take the course — see Team Participation.

    Why take it

    • Speed. Take on larger-scoped projects in the same amount of time, and finish scoped analyses in days rather than weeks.
    • Leverage. Automate the repetitive parts of your work with systems you build and control — not black boxes.
    • Trust. Learn to tell quality from AI slop, validate agent and model output, and keep scientific standards high while moving fast.
    • Reach. Ship tools and reports that non-technical stakeholders can actually use, connecting your science to impact in other domains.

    What you'll be able to do

    By the end of the course you should be able to:

    • Produce a polished project report in collaboration with Claude Cowork.
    • Use coding agents for scientific programming — and validate their output.
    • Run and orchestrate multiple parallel agent workstreams.
    • Build reliable, reusable agentic skills.
    • Build a REACT loop and a full agentic workflow, connect it to data, and validate, monitor, secure, and launch it for yourself and others.
    • Architect, build, train, and tune deep learning models for your own scientific problems.
    • Build on foundation models, evaluate them rigorously, and integrate them into research pipelines.
    • Judge when a deep learning or foundation model is the right tool — and when simpler methods win.
    • Tell the difference between quality and AI slop — and engineer for quality.

    How the course is organized

    Four weeks, each with a core focus:

    1. Week 1 — Agentic Cowork Tools (July 20–24, 2026). Research, report, and build with Claude, Claude Cowork, and Claude Code. Open to everyone.
    2. Week 2 — Agentic Systems (July 27–31, 2026). Build and launch your own agentic workflow, from a first REACT loop to a monitored, secured, production-ready system.
    3. Week 3 — Deep Learning (winter 2026–27, dates TBD). Architect, build, train, and tune deep learning models.
    4. Week 4 — Foundation Models in Research (winter 2026–27, dates TBD). Build on the deep learning core: evaluate models, build on other models, and incorporate them intelligibly into your research.

    See each week's schedule in the sidebar for day-by-day timing and topics, and the Setup & Day 0 Checklist for what to do before July 20.