Building and Enlightening Data Professionals in Africa.

An annual conference for all data practitioners in Africa.

#DataFestAfrica22 #DFA22

The path to ML mastery and it's not “what you are thinking.”

Steven Kolawole

ML Collective | Nazari

@steveddev

An Annual conference for all Data Practitioners in Africa.

“What you are thinking”…

An Annual conference for all Data Practitioners in Africa.

• Take Python, data science, and machine learning courses and, maybe math classes too.
• Do Kaggle and Zindi competitions to build your machine learning skills.
• Write technical articles. Try to teach others.
• Learn software engineering for model deployment and DevOps for performance monitoring.
• Build personal projects. Maybe build an E2E ML application incorporating data engineering and MLOps.
• Contribute to open-source.

An Annual conference for all Data Practitioners in Africa.

It’s Research!

Profile Summary

- ML Researcher @ ML Collective

- ML Engineer @ Nazari | ASAlytics

​

​

• Low-budget philosopher
• Clowning-as-a-Service

Steven Kolawole

An Annual conference for all Data Practitioners in Africa.

Photo comes here

I know a few things about Python.

An Annual conference for all Data Practitioners in Africa.

Photo comes here

I know a thing or two about taking Python/ML/Maths courses.

An Annual conference for all Data Practitioners in Africa.

Photo comes here

I know a few things about building ML projects, open source dev, and winning hackathons.

An Annual conference for all Data Practitioners in Africa.

Photo comes here

I know a few things about ML Research.

Photo comes here

What I assume about you

A couple of Caveats

An Annual conference for all Data Practitioners in Africa.

A couple of Caveats

• I am not a top 1% in all the paths in “what you’re thinking” so some of my hot takes might be wrong.

An Annual conference for all Data Practitioners in Africa.

A couple of Caveats

• I am not a top 1% in all the paths in “what you’re thinking” so some of my hot takes might be wrong.
• It’s not possible that all my hot takes would be wrong.

An Annual conference for all Data Practitioners in Africa.

“What you’re thinking” #1:

Taking courses

An Annual conference for all Data Practitioners in Africa.

Why I like courses

• A very good way to start out your ML journey.
• Helps you hone specific skills in more advanced topics.
• Instils standard practice - super-useful when just starting out.
• Courses help you gain momentum.

An Annual conference for all Data Practitioners in Africa.

But…

An Annual conference for all Data Practitioners in Africa.

Photo comes here

Momentum will stop momentum-ing 🤌🏾

Tutorial Hell is real and it's one of humanity's worst known sufferings. You can easily find lots of noobs burning, out there, in the fire of tutorials.

An Annual conference for all Data Practitioners in Africa.

The law of Diminishing Returns

• ML Courses are usually high-level. Great for newbies, not great for non-newbies.
• The low-level ML courses are usually too “boring” or “dense”.
• Huge gap between courses and real-life.
• Courses, like meetings, create an illusion of being super resourceful.
• You’d probably forget most of what you learned in that course in 6 months. Or less.

An Annual conference for all Data Practitioners in Africa.

Photo comes here

You most likely won’t be able to function in a junior ML role after finishing the most popular ML course in the world.

​

​

Can go without saying but…

An Annual conference for all Data Practitioners in Africa.

“What you’re thinking” #2:

​

Participating in Kaggle/Zindi Competitions

An Annual conference for all Data Practitioners in Africa.

Photo With Title

Photo With Title

Photo With Title

Photo With Title

Photo With Title

“What you’re thinking” #3:

​

Write technical articles. You become a master by teaching others.

An Annual conference for all Data Practitioners in Africa.

Photo comes here

By teaching we learn, but,

An Annual conference for all Data Practitioners in Africa.

Photo comes here

By teaching we learn, but,

• It is super easy to get stuck in the reward loop and forget the reason for towing this path in the first place.

An Annual conference for all Data Practitioners in Africa.

“What you’re thinking” #4, #5 & #6:

​

​

Spend time on Software Engineering

An Annual conference for all Data Practitioners in Africa.

The many pros of SWE in your ML trajectory

• Your models are not just ending up in IPYNB files,
• You become a better & well-rounded programmer and versed in standard practices as a result,
• Building E2E ML apps will always stand out on most CVs,
• Contributing to open-source will feel like working in a top-tier firm, all points checked,
• If you work in an AI/ML startup, SWE will be important as you’d need to be a generalist lots of time,
• etc.

An Annual conference for all Data Practitioners in Africa.

But…

An Annual conference for all Data Practitioners in Africa.fb

• Tendency for knowledge base to increase in breadth while lacking depth.

An Annual conference for all Data Practitioners in Africa.

Tendency for knowledge base to increase in breadth while lacking depth.

ML Interviews:

An Annual conference for all Data Practitioners in Africa.

• Tendency for knowledge base to increase in breadth while lacking depth.
• Tendency to switch to SWE completely.

An Annual conference for all Data Practitioners in Africa.

Tendency to switch to SWE completely.

So why Research?

Photo comes here

• Keep up with the state-of-the-art in machine learning.

An Annual conference for all Data Practitioners in Africa.

Photo comes here

• Understand the building blocks of the ML tools you use regularly.

An Annual conference for all Data Practitioners in Africa.

Photo comes here

• Your interests in ML is extremely limited without exploring research.

Photo comes here

• Reproducible science: Open source documentation and standard SWE practice.

​

​

An Annual conference for all Data Practitioners in Africa.

Photo comes here

• You might enjoy research enough to transition from an ML Engineer to a Research Engineer.

​

​

An Annual conference for all Data Practitioners in Africa.

In conclusion…

​

• You don’t need to be sophisticated with ML before you start reading/implementing research.

An Annual conference for all Data Practitioners in Africa.

In conclusion…

​

• You don’t need to be sophisticated with ML before you start reading/implementing research.
• No dulling, chiefs! Start now!!!

An Annual conference for all Data Practitioners in Africa.

Thank you Sponsors

An Annual conference for all Data Practitioners in Africa.

THANK

YOU!

Steven Kolawole

ML Collective | Nazari

@steveddev