Andrew Ng’s 5 Crucial Mindset-shifts for Transitioning Into the Industry From Academia

It doesn’t have to be confusing or hard if you listen to the experts

Andrew Ng’s 5 Crucial Mindset-shifts for Transitioning Into the Industry From Academia
Photo by Oliver Sjöström from Pexels

I’ve been questioning my future in data science for a while now.

I know for sure I want to be in the data science and AI space, but should I pursue a Ph.D. for an academic career or stick to what I know the best — an industry-focused data science career?

So I spoke to a few experts who transitioned from academia to industry and vice versa.

The path to pursuing an academic role seemed straightforward: enroll in a Ph.D., gain in-depth expertise in one of the fields, and continue the academic career (I’ll keep this for another day). However, transitioning from academia into the industry isn’t straightforward.

You need to know how to translate and solve business problems through data science; accuracy might not matter as much as the time to deliver results. Communicating results to a stakeholder is a whole different ball game.

Interestingly enough, Andrew Ng has also written about it, and what follows is my understanding after talking to a few experts and his thoughts.

1. Accuracy or Speed: Which One to Focus On?

The rest will come to you naturally if you can adopt this mindset.

In research, when you compare algorithms, accuracy (or any performance metrics) is crucial. If a researcher claims that an algorithm is better than a previously published work, it is better to be correct. The researcher would be required to perform experiments and showcase the results. The time taken to develop the algorithm hardly matters.

In the industry, often, there is no correct answer to the accuracy vs. speed question.

An algorithm that is accurate 95% of the time, built in a year may not be better than solving the problem with 92% accuracy in three months. Businesses certainly prefer and may like a solution that is acceptable and available in a shorter time which can be improved in iterations.

“When can we see a prototype version of this?” — is often a question I hear whenever we pitch an idea to senior management. Most often, quicker solutions with resealable performance that can be iterated later on are your best bet if you can manage the expectations of the stakeholders.

2. From Novelty to Return on Investment

My previous manager was a Ph.D. holder, and his eyes always lit up when he heard a fresh idea. He’s a hardcore academic wired to seek novelty in everything we do.

Sadly, he feels disappointed when most of these fresh ideas get turned down by the senior management. It is because the corporate world evaluates innovation in terms of return on investment.

What are the costs of working on the idea? What and when can the returns be materialized? What are the risks involved? These are some of the questions often asked. More often than not, an older idea with better execution profits the company.

This is not to say that novelty has no place in the industry; it has. We also need to be vigilant about the return on investment while pursuing novelty and innovation.

3. Hire Experienced Professionals Too to Build an Outstanding Team

Academia and research labs can transform students who don’t know how to code into publishers with excellent research. Senior researchers and professors have the utmost pride in shaping a student’s progress and have the right to feel so.

However, due to the expected execution speed and large teams, there’s not sufficient time to upskill the entire team from scratch by a few individuals.

According to Andrew, corporate managers tend to hire junior teams even when the team requires senior professionals with sufficient expertise. Thus, he recommends hiring experienced people alongside educating the existing team to keep up with the expected delivery speed.

4. Become Comfortable Working on Domains Outside Your Expertise

In academia, it’s normal to focus on a niche domain and specialize in it. The further you specialize, often better your research contributions would be. Most academic professionals are known for a particular domain and are the team’s go-to person for that domain.

In the past year, I’ve worked on three domains in the data science industry. For each problem, I had to work with different business knowledge experts, gather their knowledge, and apply it to the problems we were trying to solve.

Often in the industry, you must collaborate with different teams, such as product development, software engineering, sales and marketing, and community. It’s vital to have an understanding of domains outside of your expertise and be able to explain domain-specific topics to other teams as well.

Solving data science problems is teamwork, and having an open mind about others’ expertise and willingness to learn is the key to success.

5. Embrace Top-down Management

Yes, we’ve heard much about the flat culture and independent teams, yet the company runs on top management’s vision.

To be honest, there’s nothing wrong with top-down management. The vision and goals are set, and it’s expected that the teams are aligned on it and execute it successfully.

However, in an academic setting, top-down management rarely exists. Most decisions are made at the individual or group level as they advance in the research. The researchers are primarily in control, and the results are communicated to the stakeholders.

So when transitioning from academia, it’s crucial to adopt a mindset where some decisions and goals might already be taken for you, and you have to play a role in executing them. There’s nothing wrong with either approach, but understanding the difference in advance can help transition smoothly.

Final Thoughts

I’m not here to convince you to move into the industry from academia or research labs. I haven’t worked in academia long enough to compare academia vs. industry and pick one. I want to do full-time research at some point in my career.

Andrew Ng and I agree that if you choose to move to the industry from academia, there are five fundamental mindset shifts that you need to work on to make the transition smooth.

  • Not losing speed while searching for accuracy and other performance metrics.
  • Understanding the return on investment for each project
  • Hiring experienced teams in addition to junior teams
  • Focusing on interdisciplinary work
  • Embracing the top-down management

We need academia to spearhead the innovation and novelty in the field and industry to pioneer the adoption and implementation in practical scenarios.

For more helpful insights on breaking into data science, honest experiences, and learnings, consider joining my private list of email friends.