Glasseye Read the newsletter

A Glasseye course

The New Science

Agents have changed the work of data scientists – there's no going back. Used correctly they free you up for the proper work of a scientist: testing theories, building tools, and, above all, solving problems. Learn how to make agents work with you to produce robust, effective data science.

  1. IThe new environment
  2. IIThe new tools
  3. IIIThe new skills
Sign up now

Almost everyone is saying the same thing…

The first 80% of an agent-run data-science project is magical; the final 20% a nightmare of unpicking errors and sifting through reams of unverified output.

The problem is not with the tool, but with how it is used. Agents used with self-discipline, the correct level of oversight, and an understanding of what they should and shouldn’t do are phenomenally effective.

In other words, it’s all in the how. This course will introduce you to the new environment, familiarise you with the new tools, and teach you the new skills. You will leave with an understanding of what it means to be a data scientist in the post-GenAI world.

The syllabus

I

The new environment

Session one

In which we explore in detail:

  • What has changed for data science as a result of generative AI
  • Why we must adapt to these changes
  • And how we should adapt

The “how” question is of particular importance. Here we dissect:

  • What data-science teams are getting wrong in their current approach to the use of agents and generative AI
  • How to stay in control, maintain integrity, and continue to learn when working with agents

Finally we will look at strategies for succeeding in a tech environment that is rapidly evolving.

II

The new tools

Session two

In which we introduce you to:

  • The new tools and concepts – not just coding agents, but skills, MCPs, connectors, plugins, hooks, harnesses, evals, guardrails, and the use of multiple agents
  • Old tools to be used in new, very different ways: CLIs, linters, validators, unit tests, notebooks, config files, repositories, versioning software, graphs, and knowledge bases
III

The new skills

Session three

This is where it all comes together as we learn how to effectively deploy the new tools in the new environment.

This means:

  • Going back to the fundamentals of science – coding agents give fresh relevance to:
    • Preparatory analysis, including mathematical modelling, ontology specification, and concept clarification
    • The scientific process of conjecture and refutation, with the agent as critic/adversary
    • The need for clear and precise prose, at the right level of detail, when writing prompts and instructing agents
    • Managing teams, human and virtual, in disciplined pursuit of a research goal
    • Quality control through guardrails, testing, and the management of uncertainty
  • And rethinking existing disciplines: statistical modelling, machine learning, and causal analysis

Finally we present a fully worked example of a data-science project using the tools and skills.

Why do this with Glasseye?

The course is run by Simon Raper, a veteran data scientist with 25 years of experience of problem solving in data science, statistics, and machine learning. He has trained and mentored hundreds of data scientists, including at Google, The Economist, and ITV.

Simon is also a rabid early adopter of new technology and has been adapting his methods since the early days of generative AI. He has both the old knowledge and the new techniques.

The Glasseye newsletter, written by Simon, fiercely promotes good practice in a sector dominated by hype, and explores the implications of the generative-AI revolution for data science, machine learning, and statistics.

Join The New Science

The course package

  • Three 90-minute online workshops
  • A 30-minute one-to-one technical development session with Simon
  • A one-year subscription to the paid tier of Glasseye
Sign up now