Master Agentic Data Science
A live workshop for data scientists, analysts, and engineers who want to 10x their output with AI agents, without sacrificing statistical rigor, reproducibility, or the credibility of their work.
AI builder, consultant, educator of 6+ million students; ex-Yale, ex-Max Planck.Thomas Wiecki, PhD,
Co-Author of PyMC, founder of PyMC Labs. Former VP of Data Science at Quantopian.Luca Fiaschi, PhD,
Partner, PyMC Labs. Former Chief Data & AI Officer at Mistplay, VP at HelloFresh.
The challenge
AI tools are creating a new kind of leverage,
and a new kind of risk.
Data practitioners can now produce sophisticated analyses in a fraction of the time. But without the right structure, that speed creates confident-looking work full of hidden errors. The difference between the two is knowing how to direct AI rigorously, not just quickly.
You're still doing it the slow way
Hours lost to EDA, cleaning, and boilerplate that AI could handle in minutes. Meanwhile, your actual expertise sits idle.
AI-generated analysis is a liability
Models confidently hallucinate statistical conclusions. You can't catch errors you don't know to look for, and by the time a stakeholder spots them, the damage is done.
Prompt-and-pray isn't a workflow
No structure. Nothing reproducible. No way to scale what works or hand it off to someone else.
Direct AI through the full analytics stack
From raw data to Bayesian decision-making. Spec-driven, reproducible, and something you can actually defend in a meeting.
Catch statistical errors before they ship
Adversarial multi-model review and confounding checks baked into every pipeline. Errors get caught before they leave your notebook.
Build systems that compound in value
Reusable Agent skills and open-source code that make your next analysis take hours, not weeks.
Five skills you'll use the week after.
Not theory. Not toy examples. By session 4 you'll have operational systems running on complex business datasets.
Use AI for advanced data science
Descriptive, causal, predictive, and Bayesian modeling - applied to a real business dataset. End-to-end workflows executed in minutes, not weeks.
Engineer effective context and skills
Define the spec, review it, then build. No aimless prompting, just reproducible results from day one. Be clear about when to use LLMs coding capabilities and when to use deterministic scripts.
Produce error free results
Catch what AI misses - statistical errors, confounding, misinterpretation. Stress-test outputs with adversarial agents and enforce quality control gates.
Embrace uncertainty in decision-making
Move beyond point estimates to Bayesian thinking. Quantify uncertainty, understand each LLM strengths and limitations, and choose the right model for the problem at hand.
Scale your workflows exponentially
Build reusable templates and parallel pipelines. Turn weeks of analysis into minutes - consistently.
Three instructors. Three different angles.
Most courses pick one thing: theory, tools, or application. We didn't want to pick.
Independent data and AI consultant. Has taught data science & AI to 6M+ learners. He has advised and taught teams building AI-powered systems, including engineers from Netflix, Meta, and Amazon. He’s the one who makes sure you actually build something that works.
Co-author of PyMC, the most widely used open-source Bayesian statistics platform. Runs PyMC Labs. Former VP of Data Science at Quantopian. PhD in Computational Cognitive Neuroscience from Brown. He’s the one who makes sure your stats hold up.
15+ years running AI and analytics teams at fast-growing companies. Former Chief Data & AI Officer at Mistplay, VP of Data Science at HelloFresh. Built experimentation platforms and drove real revenue with causal modeling. PhD in Computer Science from Heidelberg. He’s the one who connects the model to the business.
You get all of them in every session.
Four sessions, learn to solve real business applications.
Each session teaches both the data science concepts and agent skills you need to apply effectively in real business use cases.
One price. Everything included.
Two weeks. One price. You walk out with a working agentic data science workflow.
- ✓4 live sessions (12 hours of instruction)
- ✓All course materials, datasets & lab notebooks
- ✓Reusable agentic system you keep forever
- ✓Session recordings (lifetime access)
- ✓Certificate of completion
- ✓Private community channel & alumni network
- ✓Post-cohort office hours with instructors
What's one month of your time actually worth?
Do the math: if you save even 5 hours a week, the course pays for itself in two months. And the templates you build in session 4 stick around. Teams of 3+? Email us about group pricing.
What past students said
From data scientists who've taken previous PyMC Labs + Vanishing Gradients workshops.
I've been applying the lessons from the course to how we build observability and evaluation for our AI agent.
Frequently Asked Questions
Quick answers before you commit.
You should be comfortable with Python and have some machine learning experience (scikit-learn level). The course is for people who want to work faster without cutting corners on quality. If you are looking for shortcut prompt tricks or fully automated AI, this is not for you.
You should be comfortable with pandas, NumPy, and scikit-learn. No PyMC or Bayesian experience needed. Thomas introduces that in session 3.
Copilot gives you code completions. ChatGPT gives you one-off answers. This course teaches you to build structured, reproducible workflows where agents work together and you catch their mistakes. It's the difference between a tool and a system.
Each session runs 3 hours (Mon & Wed, 3–6 PM Pacific). Add roughly 2–3 hours for optional assignments and review. Most participants budget 5–6 hours per week across the two-week cohort.
All sessions are recorded and posted within 24 hours. TAs are on Discord for async questions, and there are weekly office hours if you need to catch up.
A working end-to-end agentic data science workflow. A library of reusable templates. A capstone project for your portfolio. And access to the PyMC Labs alumni community.
Yes, team pricing is available for groups of 3 or more. We've run workshops for teams at Roche, Bain & Company, and Colgate-Palmolive. Email [email protected] for a quote.
Yes. You get lifetime access to all course materials, including session recordings, notebooks, datasets, and templates. They’re yours to keep and revisit whenever you need them.
Yes, you’ll need access to an AI coding agent such as Claude Code, Cursor, Amp, or similar. It doesn’t matter which one you use—the workflows we teach are agent-agnostic. We’ll include setup instructions in your confirmation email.
As a consumer, you have the right to withdraw from your enrollment contract without giving any reason within 14 calendar days of the date of registration ("withdrawal period"). This right applies provided the course has not commenced prior to the expiry of the withdrawal period. Refunds will not be available once the course has commenced. This right of withdrawal applies exclusively to consumers and does not extend to individuals enrolling for business, professional, or commercial purposes.
Ready to do data science at AI speed?
We keep cohorts small so every session stays hands-on and you get real time with the instructors.
Registration is currently closed.
We're not enrolling right now. Registration will reopen once the next cohort is confirmed — check back soon.
Registration Closed
Enrollment for this course is currently closed. We'll reopen registration as soon as the next cohort is confirmed.
