There is a lot of hype around ML. We will overcome it and immediately build a real model that is useful for business, and at the same time, define the basic concepts of ML problems.

You will learn:
→ That it is possible to build an ML model in Excel
→ Main terminology of ML: feature, target, training set, model

What it looks like:
→ You help the CEO solve the problem of demand for rental bicycles
→ Using different industry cases, you practice to determine what features, objective function, and training set are.

Although PMs do not need to understand the internals of algorithms, they need to choose from a variety of problems those that are more efficiently solved using ML, and the ones that can be solved the old-fashioned way or bypassed altogether.

You will learn:
→ Main types of ML problems (regression, classification, ranging, etc) based on the examples from the domain
→ What tasks should be avoided
→ What problems do not require ML at all

What it looks like:
→ You help your colleagues (CFO, Marketer, Jr.PM, CEO) to determine their problem type

When the PM sees that the problem belongs to one of the known categories, it is extremely important to understand what critical data we will need to build a fast MVP, and where to obtain this data.

You will learn:
→ How to identify features for training the first version of the model
→ Five sources which you can get training data from

What it looks like:
→ You help your colleagues to identify the main features for their cases
→ You come up with ideas where to get data for these features

When the PM sees an ML task that is useful for business, selects the first features, and even gets hold of training data datasets, it's time to transfer the scope of work to Data Scientists. Then the magic happens, and the PM returns to the discussion when it's time to talk about model quality metrics (see next lesson). But what kind of a PM doesn’t want to also understand a little about magic?

You will learn:
→ What steps a DS takes when building a model
→ Where to get test and training data sets
→ How the model becomes more complex when it is necessary to take into account more complex dependencies

What it looks like:
→ Together with the Data Scientist, you look at the data and Python code of the bicycle demand model
→ You complicate the model so it becomes a nonlinear one
→ And complicate it even more, so you have two variables now

As with product metrics, the PM does not need to be a master of detail, but it is important to understand the output of the trained model and be able to discuss it with the analytics team.

You will learn:
→ That the Data Scientist will usually give you not one, but several versions of the model
→ What is Recall, Precision, and F1 score for the classification ML model, and why we need them
→ What is RMSE and MAR for the regression model and why we need them

What it looks like:
→ You learn to calculate Recall, Precision, and F1 score based on cases presented
→ Them, you choose which version would be better based on your understanding of this particular business case

It's time to apply all the knowledge gained to a new task from our CEO.

You will learn:
→ How to apply the acquired knowledge from start to finish

What it looks like:
→ You solve the problem of predicting user LTV

We will discuss how ML skills relate to other product skills: analytics, technology and A/B testing

You will learn:
→ How ML relates to API/services, A/B experimenting, and analytics
→ How to grow further in the domain: the PM AI role

What it looks like:
→ You receive a summary checklist and links for further study
→ You receive a personal certificate