In short
Data science is not the right move just because it is fashionable, argues Alka Pandey, Manager Data Scientist at Schneider Electric. The real foundations are statistics, math and solid programming, and the field is wider than model building: data engineering matters just as much. Everyone benefits from basic data literacy, but only those genuinely interested should change careers to pursue it.
Key takeaways
- Understand the industry landscape of data science.
- Understand the jargon in the data science industry.
- Decide whether you should invest time in learning data science even if it doesn't concern you at the moment.
- Learn how you can transition into a data science career track.
Start with interest, not the hype
Pandey opens by acknowledging the buzz: everyone, it seems, wants to move into AI and machine learning. Many people ask her how to make that transition, and her answer always starts in the same place. "The first question I asked them, are you really interested in this? Because it's not about getting into this field just because it looks great." (0:58)
She is equally clear that some data knowledge is now unavoidable for everyone. "Data science is now everywhere. You cannot escape data science." (1:57) For those who do want to go further, she points to online courses on Udemy or Coursera, including those by Andrew Ng, or a one-year full-time programme such as an MS or an MBA in business analytics.
The jargon, decoded
A large part of the talk strips the mystery out of the vocabulary. Pandey offers a deliberately simple definition: "So whatever you do with data is data analytics. That's the simple definition of data science and data analytics." (0:30) Business intelligence, data mining and deriving meaningful information for stakeholders all fall under that umbrella. What has changed, she says, is that machine learning and AI are now part of data science too.
Artificial intelligence gets an equally plain explanation: "AI is nothing but how you can make the machines talk and think like humans." (0:58) Her examples are everyday ones: Alexa, Siri, Cortana, Tesla and self-driving cars. Deep learning and neural networks, she adds later, are more advanced forms that help machines model the structure of the human brain.
Why does all of this matter now? As the speaker cited, people were generating 2.5 quintillion bytes of data, and that figure has since grown. The volume, velocity and variety of data are rising so fast that organizations need tools to make sense of it and drive business decisions.
Statistics and math come first
For anyone serious about the field, Pandey's advice is to begin with fundamentals rather than tools. "Basically the backbone of all machine learning and AI is statistics." (4:15) She lists the starting points: hypothesis testing, p-values, inferential statistics and descriptive statistics. Descriptive statistics becomes exploratory data analysis in data science, and predictive analytics is, in her words, predictive statistics.
The math list is concrete: algebra and calculus, straight lines, linear and nonlinear equations, eigenvectors, determinants, matrices, probability distributions, and measures of central tendency, dispersion and variability. Domain knowledge and analytical skills complete the picture of a good data science professional.
Myths and facts
Pandey tackles three common myths. The first is that you can be a data scientist without programming. Not true, she says: "You have to know, in fact, rigorous study and understanding of the subject is required for data science." (6:04) Data engineering, data science and data analysis work has moved to Python or R, and whichever language you choose, you need to be good at coding.
The second myth is that a degree is unnecessary. She calls this partly true, but notes that companies are looking for certified data scientists, and a degree shows you have spent enough time on the subject, so she recommends one for newcomers. The third is that data science and statistics are separate fields. They are not, which is why people with a statistics background often find data science a natural fit.
Hype, reality and the rise of data engineering
Pandey accepts that data science is one of the most sought-after job profiles, but warns against chasing trends. "So don't run behind technologies. Do what you love." (7:16) If you love management, do that; if you want to be a singer, sing.
She also redirects attention to a less glamorous role. As the speaker cited, current trends suggest data engineering is also well paid, and she expects demand for data engineers to exceed demand for data scientists within two or three years, because data scientists need data before they can do anything. "Ignoring it and just going to the machine learning or model building is probably not right." (8:28)
She closes with a tour of applications: fraud detection in banking, dynamic pricing in railways and aviation, recommendations on Flipkart, Amazon and Netflix, inventory planning, delivery route optimization and healthcare research. Yet she draws a firm line: "So know the differentiation between knowing something and transitioning your career into those kind of a field." (10:39) In the brief Q&A, she adds that deep learning, neural networks and NLP are what currently catch employers' attention.
Watch the talk
De-Mystifying Machine Learning and Artificial Intelligence, by Alka Pandey (13 minutes): https://www.womentech.net/video/de-mystifying-machine-learning-and-artificial-intelligence
About the speaker
Alka Pandey is a Manager Data Scientist at Schneider Electric, solving complex supply chain problems with data analytics, machine learning and artificial intelligence. She is also a data science trainer whose goal is to make data analytics and machine learning more intuitive and easy to understand for everyone.
Questions this talk answers
- Do you need programming skills to be a data scientist?
- Where should a beginner start learning data science?
- Do you need a degree to move into data science?
- Why does data engineering matter so much?
- What skills attract employers in data science?
Drafted with AI from the talk transcript. Editor review pending.