8 Habits of Successful Data Scientists

After watching and working with successful pioneers in data science and machine learning, here are the everyday habits I have seen

I am very lucky to meet some pioneers of AI in my career. Their viewpoints shaped my current interests and thinking, and their habits formed my daily routine. I often get asked, mostly when I speak in conferences, ‘How do I become a successful data scientist’. And my answer is always — it starts with forming some crucial habits. Here are the 8 practices specifically that worked for me and made people's exceptional data scientists.

Optimize your Workspace and Toolkits

I am quicker than most professionals because I have streamlined the tools I use. Since I do not waste any fighting with my development environment, I always deliver projects sooner. I know my AWS environment. I have found out where to use which programming language and database(s). I have scripts ready to clean my datasets often, build models and hyper tune parameters. Whenever I start some new project, I don’t start from scratch. Consider this, it took me 1 year to build an automl product, but now that kind of automation saves many other professionals and me thousands of hours!

It took me a great deal of experimentation to arrive here. I’ve used many designs than I can tally and explored different data sciences and development applications. You should develop what works best for you and enable you to be more productive and effective.

Focused Reading & Learning

You can only read and consume content only for a couple of hours at maximum. There is more trash than gold to read. Build a list of trusted newsletters, blog aggregators (my favourite being towardsdatascience) etc. Try not to waste time reading the fluff articles like on Forbes. Don’t read every in-depth learning paper you get your hands on and neither try to understand every open source code on github.

I have subscribed to newsletters which bring me research taking place in companies like FANG, Tesla etc. I follow many of their executives on Twitter, Linkedin, to keep updated about AI work and research.

Stanford, MIT etc are great source of strong quality search. Find bloggers who write/talk about topics of interest to you in a format that is easy for you to comprehend. Everyone writes and speaks about similar themes differently, so pick your style of communication/presentation styles.

Build A Professional Network

Reputation and influence are the new marketing. Two advantages make the work justified. Firstly, you can only learn from your professional network. For instance, if I need some brains to pick on while solving a problem, I call in the experts. Secondly, a professional network allows you to build a brand and grow your influence in the AI community.

Twitter, Linkedin are great platforms to build a reliable professional network. I will be soon writing a blog on how I grew from 1k to 6k connections in a year (stay tuned). You can add yourself to various telegram and slack groups as well and connect with like-minded people. I recently started one to connect industry leaders with data scientists. Start by following and listening. When you figure out the content that the community likes to engage in, become an aggregator. Aggregating merely is sharing content that your connections will appreciate as well. Find your voice in the groups. See something interesting that doesn’t go into details? Have a unique perspective or something interesting to share? Create content! Push your content on Linkedin, blogging websites, Quora, Reddit etc. Hold sway. Start answering questions on StackOverflow and others. As you acquire mastery, consider adding to open source projects or publishing your research.

Influence brings opportunities. Rather than cold calling for new jobs and promotions, you will constantly be promoting yourself. I get speaking opportunities, chance to be on panels, insider access to conventions and many new clients through my network. It is worth the effort.

Have an ear for Business Problems

Just want to clean data and build models on top of it? I have some bad news for you, people like me are automating feature engineering, model building, tuning parameters and even deployment. So, you need to find new business problems to solve and have an ear for business problems. I listen to NPR, Bloomberg Business, CNBC and read WSJ every day. What I’m listening up for is the reason organizations missed profit targets. Those are the questions they have not solved and need answers and will pay for solutions. Also, these platforms evolve my thinking and give me multiple things to ponder on.

How can AI tackle these challenges? I see a tonne of predictive problems; something happened that the organization didn’t foresee. It could be a change in customer behaviour/preference or something like supply chain disruption. I additionally see machine learning capabilities issues; the management is not data literate. They have no clue what to do with data. I consult businesses, and I have seen over 60% of companies at maximum using BI. Figure out how to solve these problems and publicize your capacities. You are much more significant to the company if you can understand their business problems come up with solutions.

Adopt A Minimalist Style

The experts use the fewest lines of code, the least information/data, the most straightforward algorithm, speak briefly, and so on. Minimalism is the mark of an expert data scientist.

Learn to say No (a lot)

Data Scientists are magicians! We can do a lot of things. That prompts us getting requested to do various things. Trust me in most cases; no is the right answer. It doesn’t mean you don’t like doing something, but rather it comes down to where you want to allocate your efforts and time and choosing your professional path. I have said no to jobs, projects that did not align well with my career/ambitions/personal path. Today, I am very happy and focussed on right things.

image by author

Look at this Venn diagram. The overlap is minimal. Very few opportunities will land in that overlap for you. For me, it’s tiny. The chances I need, I should pursue myself. I’m the one requesting them, not the opposite way around. The best AI practitioners I know get their projects, roles and clients. I copy that conduct myself, and it’s turned out great for me.

Take out the time and make efforts to Speak.

Grab every chance to speak. Speak at conferences, team presentations. A big reason, I have been called an influencer is because of my speaking arrangements, and surprisingly, I have been a big introvert whole of my life. But I took steps and efforts to break out of my comfort zone. Currently, I take small, private audiences that I can manage. I prepare topics in advance and do diligent homework before going to speak. While speaking, I spend much of my time listening and answering questions. We all have different styles; mine is to understand my audience’s interests and move conversations around it. I also like to involve the audience while answering and actively invite them to answer or have their say if they think they are experts in that question. When the audience doesn’t have questions, it makes more sense to give a presentation, kind of guided tour. However, this is better suited for a conference setting.

Come up with your style, the message you want to give and what type of audience you want. Talk about your unique perspectives, experiences, projects and visions. You’ll be astounded by the number of individuals who are keen on what you can share and instruct them. Your viewpoint is definitely more important in your long AI career than some other commitment you’ll make in code or calculation.

Speak with a purpose and a goal to be understood

I spend just about 20% of my time speaking with data science specialists. Most of my exchanges are with non-technical crowds. They couldn’t care less about AI. They have an outcome they need. Data Scientists who stand out explain their projects in money and math. ARR, MRR, Annual Savings, and Margin. It must be part of your vocabulary. Those are the model accuracy metrics that matter to the business. They understand ROI, not AUC.

Coming back, Bernard Baruch once said, “Most of the successful people I’ve known are the ones who do more listening than talking.”. It would be best to ask questions to stakeholders and then listen to them to understand what they want. It is an art, which unfortunately is not taught in school and most data professionals suck. A reason why, we are clubbed together with the IT team, whereas we are making strategies for the companies. I have seen the experts dissect a problem down to its root by asking the right questions

The lion’s share of speaking with purpose and clarity is first listening. The process of asking questions to get to the truth of what a person or group needs is an art form. I’m still working on this myself, but I’ve seen the masters dissect a problem down to its root by asking the right questions. The components of a decent argument are establishing a climate where individuals are open to addressing questions sincerely and conceding what they don’t know, assuring them that investing the time to answer these problems will benefit them — blending their answers back to them such that it shows cognizance.

When I comprehend the question, I can answer it with a lot more surety. I do not use technical jargons, I am not looking to impress anyone with my model’s ROC. I use simple business English to explain complicated concepts. For instance, there is a neural network waiting to be explained, I will use a decision tree to explain that. It is on me to express ideas and concepts so that my audience, stakeholders, can understand. I do not want them to walk out of meeting all confused. I always make sure they have 3–5 key takeaways and recommendations. I call them easy to swallow pills because these points stay in everyone’s mind, leading to many more discussions and implementation eventually. This is the piece of my recommendation I discover most challenging to take; however, that perspective on communication has helped me improve significantly.

If I have missed out on something, please do write in comments, I will try to include them and share your thoughts. I am always about listening. :)