Photo by Aaron Burden on Unsplash

Approaching ML — How to start?

In a previous article, I told how I started to approach ML but, of course, my example is not good enough to generalize: available time, skills, expectations, background are all factors to consider and can have a huge impact while choosing an approach.

Moreover, talking with different persons, with different roles and backgrounds but all interested to deepen knowledge on these themes, I realize there is a great curiosity but doing the first step could be difficult and suggest something (a book or a course) can be ok for someone but the wrong choice for someone else.

But we need to start from some point, so let’s make a simple classification, using for example personas derived from job roles and moving from here to imagine different learning paths.

To keep it simple, let’s consider:


A person with a strong business background and no technical expertise.

He/She wants to know more about ML with the goal to answer questions like: how these technologies can improve business? At what cost? How can ROI be measured?

The CTO/Tech Manager

A profile with good degree of technical knowledge, a former developer switched to management. Could be the person appointed from CxO to make a strategic plan or start with a PoC on these topics, so his/her questions could be: what I need to know to hire right people for a dedicated team? To manage efficiently that team and projects? How to evaluate an external supplier or a cloud ML service? How to present results to top management?

The Tech

A tech person, proficient in one or several languages/frameworks with the willingness to gain new skills (and related tools) and the need to know how to implement what CxO are asking.

I know, it’s a bit simplistic but let’s consider it as a baseline.

So, how to start this journey?

My personal advice for CEO is a top down approach, meaning starting from the big picture and the essential facts to gain an high level understanding good enough to make better strategic decisions and have a common ground to choose the right persons to work with on these topics.

The amount of (technical) knowledge to achieve probably is the lowest one respecting other roles, but find the optimal trade off between generalist and specific insights could be more difficult than just having a full specific course.

At management level, it depends on several factors (specific activities, academic and working background, personal attitude to learn and to manage people) but I can say that probably you don’t need to have super low level knowledge but a solid comprehension of what can happen during a ML project is paramount, to better understand problems can arise, to interact more efficiently with the team and to help top management to answer their strategic questions.

For Tech, the matter is even more complex: the application fields are numerous with tons of frameworks, libraries, languages and specific technical goals (data retrieving, data preparation, models training and evaluation, infrastructure/operations, presentation) so the above listed factors are even more relevant.

Here’s a little motivational image :) (author: Swami Chandrasekaran)

Don’t worry, it’s not bad as it seems..first, there are different roles mixed and second no one can know everything, as no one knows every programming language and framework..

So, my advice here is to adopt the same approach working for everything both theoretical and practical: study and try things at the same time.

You can’t learn just reading books or watching videos but you need to try out and face all the practical problems that are seldom mentioned and, at the same time, just doing practical things without a good comprehension of theory maybe could work in the beginning but it’s a dead end (a fact I personally learned).

In next posts, I will focus on these aspects too, as I think overall better comprehension can lead to more adoption and help to democratize it (no strings attached here, I’m not a big cloud provider trying to sell services or GPUs time :)

Last thought: one of most famous Professor Ng quote is “AI is the new electricity”

So the question is: are you still using candles?




Tech consultant and Coach ( | Avid learner | Composer | Proudly believing less is more, except for love and knowledge

Love podcasts or audiobooks? Learn on the go with our new app.

Recommended from Medium

A History of Triggering Artificial Neuron

MuarAugment: Easiest Way to SOTA Data Augmentation

Blurring or Smoothing Out Images  —  OpenCV

Production ML Systems on GCP

Just Split, Dropout and Pay Attention

Algorithmic Trading 101 — Lesson 5: More Machine Learning

Understanding Word2vec Embedding in Practice

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Antonello Calamea

Antonello Calamea

Tech consultant and Coach ( | Avid learner | Composer | Proudly believing less is more, except for love and knowledge

More from Medium

In any batch manufacturing, understanding the variability is one of the fundamental task performed…

Data optimisation, a fine mess

Data in the Meta- Age

How should you measure the success of a machine learning model?