A Better Way To Learn Data Science Than Online Courses

This is an opinion piece. I’d love to hear your counter-arguments below.

Do you want to be a data scientist/data engineer/data analyst?

I have started my journey in January 2019, worked with two startups, one of them failed miserably and the other one is started making some decent profit recently. I did some internships and have experience in cognitive science with the defence sector. I have met 30+ data scientists over coffee in meetups and talked on LinkedIn a few over.

Here I’ll explain how these people made the transition into data science.

P.S. It was not with online courses.

1. Solve A Real Problem With Machine Learning

Pick a real problem, then solve it with machine learning.

This is hard because there is no roadmap. But it provides real experience and a story to market yourself, regardless if you succeed or fail. You may not have the ML/DL skills yet but I bet on your imagination skills, create a unique idea, at least give it as shot !!

Here are problems you could solve:

• Detect fake news
• Detecting facial recognition system (why restrict yourself to humans, research and development of animal facial recognition is a new high, google it, try them)
• Neuro-science is the basis of AI, so why just not explore it further.
• Predict home values in your neighbourhood
• Sentiment Analysis not just with social media but with something that no one thought of!
• Recommend pets to people based on their lifestyle

If your solution works (or even barely works), build a UI that others can use and post it on Hacker News or Product Hunt.

Add the experience to your resume with your title as “Data Scientist”. If it solves a problem with machine learning, no one will care that it was a one-person show.

Now you can tell this as a story in an interview, which will carry more weight than an online certification.

2. Find A Mentor Who Is An Artificial Intelligence Expert

Build a relationship with someone experienced who can recommend AI-driven solutions to problems you’re trying to solve.

This is how I broke into data science.

As a student, I don't have any idea how to break into this and then one day I met a PhD student in my neighbourhood and he recommends me to attend different meetups around the city. I started going there in the month of December 2018 and by a month I have a roadmap ready to start my journey in data science. So, I found couple of people there, who were ready to guide me.

Each week, we discussed problems and potential solutions, I worked on the implementation, then we reviewed and repeated. After 6 months we solved several important problems and the experience was invaluable.

I followed the below process to find subsequent data, science mentors.

• Message data scientists in your city on LinkedIn
• Invite them for coffee or ask for having a one-one conversation online
• Have a specific problem in mind, and your own potential solutions to get feedback on
• Follow up with results after implementing a solution

3. Do A Machine Learning Internship

Take a short-term job where you’ll be paid less but will get your hands on a real project implementing AI.

This path isn’t for everyone and works better if you’re still young or in college. Not everyone can afford to quit their job and become an intern, but you may be able to find something part-time or online.

The important part of this is getting AI-related experience on your resume.

4. Start Doing Data Science In Your Current Job

Figure out how your current company can use AI to solve a problem. Then solve it.

You may not have time between 9 and 5 while you’re at work. But if you’re motivated, do it at night or on the weekend. Then share what you accomplished.

If your current company is small, no one will argue against you adding more value. If it’s really valuable, you may be allowed to work on it as one of your day-to-day projects.

Afterwards, put the project on your resume and update your job title, if you can.

5. Do A Data Science Bootcamp

Attend a paid data science Bootcamp.( i haven't done it yet but this is on my to-do list)

This costs money and not all boot camps are equal, but I know at least 10 people who broke into data science post-boot camp, and all with large reputable companies.

The best boot camps only accept PhDs, so it is possible that candidate success relies on survivorship bias (boot camps accept students who would be successful anyway).

Bootcamps benefit candidates in several ways.

• Candidates do consulting for real companies
• Graduates are connected with companies looking to hire
• Job preparation is provided

That said, not every graduate will get a job.

6. Become A Software / Data- Engineer/ Analyst First

As long as data scientists solve problems with code, there will be heavy overlap with software engineering and data engineering process.

After gaining experience as a software engineer, find a data science job that uses a similar tech stack to your experience (ie: database, language, framework, packages).

If you can check off most job requirement boxes, you have a good chance of getting an interview.

There are other benefits to becoming a software engineer first.

• Make a decent salary
• Work at companies who hire data scientists
• Build a generalist technical background
• Prove you can do a similar type of work

7. Do A Technical PhD or Master’s Degree Then Apply For Jobs

Do you have 2 to 6 years to study? I have it.

But most of the data scientists I met followed this path.

They have either:

• An AI-related masters degree
• A technical PhD (not necessarily AI-related)

I wouldn’t recommend going back to school to break into data science. But if you’re currently in school, and can transfer to an AI-related degree, do it. Anecdotally, the highest paying AI salaries require advanced degrees.

While costly and time-consuming, traditional degrees carry a level of trust that online certification doesn’t.

What About Online Courses?

There IS a place for online courses. But it’s not for getting a job.

The benefits of courses include learning what you don’t know and deep-diving into specific techniques.

But on the flip side, courses give the feeling of accomplishment, without making you do the hardest and most sellable thing, solving a real problem.

I’d say, find a problem to solve, then use online courses to learn to solve it.

How have you seen people break into data science? I love to hear about it below.