ML for Trading in the Age of AI Agents
Stefan Jansen
ML for Trading · Applied AI · Since 2013
Run the whole ML for Trading workflow once, end to end, in a single session.
Most people meet ML for trading in pieces: a tutorial on features here, a backtest there, an agent demo somewhere else. They never join up, and the part that decides whether any of it works falls between them.
This is one continuous pass through the whole workflow, in one session, on a hundred ETFs priced daily from free data over eighteen years. You go from a market idea to alpha factors, to labels and a validation protocol, to a gradient-boosted model, to a backtest charged for commissions and spread, to the handoff to live trading. No toy data, no skipped step.
Two blocks add a thread not usually taught alongside the first. A coding agent builds part of the pipeline live, and a research agent shows how specialists reach a single forecast. The point is where an agent genuinely carries the work, and where the judgment stays with you.
The two results the session produces are deliberately the less flattering ones. The naive information coefficient carries a t-statistic of 3.02 and the HAC-corrected one carries 0.94. The backtest returns a Sharpe ratio of 0.48 with no costs and 0.44 once costs are charged. The gap in each pair is the teaching point.
What you’ll learn
You run the entire ML for Trading workflow once on real data, and leave able to tell a real result from a flattering one.
From a market idea to a costed backtest on eighteen years of real ETF data, with no step skipped and nothing simulated.
Data, alpha factors, labels, a validation protocol, a gradient-boosted model, and the handoff to live trading, in one pass.
Watch a coding agent build part of the pipeline live, in a block that is run rather than demonstrated.
Follow a research-agent trace in which specialists produce a single forecast, and see what it does not settle.
Read a naive significance test against its HAC-corrected form: a t-statistic of 3.02 against 0.94, on the same signal.
Read a zero-cost backtest against a costed one: a Sharpe ratio of 0.48 against 0.44, on the same strategy.
The session samples the trading and ML work and the agent-authoring work, so you choose a direction having done both.
You leave with the attendee repository and the recording, and can run the whole pipeline again on your own.
Workshop agenda
Beyond the cookbook: the ML4T workflow (25 min)
What the whole workflow is, why its pieces are usually met out of order, and what has to be decided before any data is touched.
From a market idea to an alpha factor (10 min)
Turning a claim about what makes money into something measurable that a model can be asked to predict.
The modern quant stack, and a first look at the data (35 min)
The tools the session runs on, and the hundred-ETF panel itself: what is in it, how far back it goes, and what it can and cannot support.
Build the pipeline with a coding agent (20 min)
A presenter-led build in which a coding agent writes part of the pipeline live. Follow along in your own agent if you have one. A 10-minute break follows.
The judgment you don't delegate: features and labels (30 min)
What the label represents, over what horizon, and against which price you could have traded. These are the decisions no agent makes for you.
Training and evaluating the model (45 min)
A gradient-boosted model on walk-forward folds, and the difference between a naive significance test and its HAC-corrected form. A 10-minute break follows.
Backtesting the strategy, with and without costs (40 min)
Scores become positions, and the same strategy runs twice: once at zero cost and once through a commission-and-spread model.
Research agents: strategies that show their work (20 min)
A recorded trace in which specialist agents produce one forecast, and what that trace does and does not establish.
Where to go next, and Q&A (25 min)
The two directions this session samples, what each one actually involves, and time for your questions.
Learn directly from Stefan
Stefan Jansen
Author, ML for Trading · Founder, Applied AI · Investing since 2013
Who this workshop is for
Anyone new to systematic trading who wants the whole picture first, before choosing where to go deep.
Developers and data scientists moving toward markets, with intermediate Python and some pandas.
Practitioners who have used a coding agent and want to see one inside a real research workflow.
What's included
Live sessions
Learn directly from Stefan Jansen in a real-time, interactive format.
Lifetime access
Go back to course content and recordings whenever you need to.
Community of peers
Stay accountable and share insights with like-minded professionals.
Certificate of completion
Share your new skills with your employer or on LinkedIn.
Maven Guarantee
Your purchase is backed by the Maven Guarantee.
Frequently asked questions
Maven for Teams
Reimbursement
Get your company to pay
Everything L&D needs: email template, receipts, and certificate of completion.
Get reimbursedTeam discount
Learn with your teammates
Save 20%+ when 2 or more teammates enroll in the same cohort.
Save 20%+ with a teamPrivate cohort
Run a cohort for your org
A dedicated cohort with a custom schedule and curriculum, tailored to your team.
Book a private cohort$299
USD
10am–2:30pm EDT