Know yourself first, what do you like, what do you want to do, then go ahead. -Avishek Nag

Avishek Nag
Data Scientist, Machine Learning Specialist, System Architect
Author of “Pragmatic Machine Learning with Python”
Bengaluru, Karnataka

Q1. Please share your educational and professional journey?

I have total 14.6+ years of professional experience in different renowned companies.  I passed out B.Tech in Computer Science from WBUT and started my career as Programmer Analyst in a reputed Software Consultancy Firm.  Later I moved to different product-based companies and gained my experience in designing complex systems. I did my M.Tech in Data Analytics & ML from BITS (Pilani) while working and started my journey of Data Science & ML from there. From the past five & half years I am working in this area and performed the role of full stack Data Scientist in multiple organizations.

Q2. What did attract you towards Data Science to become a Data Scientist?

First of all, one should have passion towards Mathematics, Probability & Statistics along with Computer Science to excel in this area. When I started my academics, AI/ML was not that much well known or well-practiced subject. In fact, I had an inclination towards Math/Stat from the beginning, but due to circumstances couldn’t study that and ended up being in B.Tech-Computer Science. However, I got an opportunity to do several projects on Pattern Recognition in Indian Statistical Institute (ISI) when I was in 3rd year. That was the beginning of getting the taste of it. But, that time AI/ML practice was so much popular across industry. It was limited to academics, research institutes and very few industrial applications. So, as usual, my career started with as normal Java developer. When I had almost 7.5 years of experience, suddenly the trend about Data Science started in industry and I also took that opportunity fill up my old aspirations to study Math/Stat. I started M.Tech in Data Analytics from BITS (Pilani) and within two years,  my journey as Data Science expert began to excel. I worked at various Data Science projects, lead the technical design activities in several organizations and solved many challenging problems.

Q3. You are a Data Scientist, Blogger, and Author. How do you play multiple roles simultaneously in your life?

Data Science is not only about doing some coding using R or Python and build models. You have to communicate the results to the stake holders. If someone is in purely academics/research, then this part may not be that much important, but in industry it is. Blogging gives me that opportunity to enhance my writing & explanation skills. It also gives you scope to explore new ideas, share your thoughts with larger audience. Whenever I got some time, I have written blogs on topics of ML. Weekends are perfect times for this. Authoring a book is far more formal than blogging. There I need to plan, prepare my subject well, decompose it into parts and then start working on it. It’s a long-term process. To start a book, I always kept 3-6 months of plan before it. Being in work, doing all these together is always tough. But, once you gain sufficient industry experience, you can develop these juggling skills. All you need is passion to write and express your thoughts. But, work should be top priority and rest of the things are nothing but results of proper planning.  

Q4. How do you handle stress in this fast and competitive life?

Don’t let the competition drive you always. Don’t really run behind it closing your eyes. To some extent, definitely competition helps to stay relevant. Know yourself first, what do you like, what do you want to do, then go ahead. Don’t always look around people and see what’s they are doing and then try to imitate the same.  It creates unnecessary pressure. You should keep your eyes open, have a plan and execute in a peaceful mind.

Personally, I like reading story books a lot, I do drawing sometimes (though nowadays I am not doing it), watch various web series.  It helps me to get relieved from stress.

I am also a regular practitioner of Math/Stat. I do study various higher-level topics of Math/Stat. In fact, I have plans to do personal research on those. Though sounds odd, but you can say, it is also one of my free time activities.  This tension-free practice of the subjects like Math/Stat without the fear of any target/exam gives you the best of it. And this power of knowledge brings the inner energy to go ahead.

Q5. You are an Author of “Pragmatic Machine Learning with Python”. Please share some brief about your book and what is special in it?

This book is for working professionals who are trying to learn ML. At least 2-3 years of work experience would be good to start with my book. Programming knowledge in Python is mandatory. It has a balanced combination of theory & practical. Along with regular topics like Classification, Regression, Clustering, some advanced stuff like Multi-Label classification, word vectors, architectural guidelines to build Machine Learning applications in real word production environment, ML contract specifications like PMML are covered there.

 Its specialty lies in covering mathematical aspect of a topic with equations & expressions as well as Python code to implement it. It creates a mapping between these two. It also has a set of solved use cases which may give an idea to the reader about how to write a story on Data Science. My book is available in Amazon and interested readers may have look at it. And so far, I have received positive reviews.

Q6. Which one skill do you like most about yourself?

Being in the industry, I have observed there are very less people at senior level who knows the end to end life cycle of Data Science applications. You will find juniors who knows the theory & modelling but they don’t know that much about how to take it to production. You will definitely find several senior folks who know how to build a system with a production-ready design but they lack theoretical ML skills or model design skills. Fortunately, I got the scope of doing both and know end to end especially at my seniority level.  This made me stand differently in the crowd.  My initial experience of software development in different product companies and passion towards Math/Stat have definitely helped me. Now I am in apposition to lead AI/ML initiatives in any organization. I know what are the potential risks and how to mitigate those.

Q7. You have made many online articles. Are online articles & courses enough to become a Data Scientist?

In short term, it works sometimes. But, in long run, definitely not. In fact, people are claiming themselves as Data Scientist just after having some online course.  This is a wrong trend. Online article/course will surely help you to get started with. You need to keep in mind that unlike other ones, this subject is heavily backed by solid Mathematical theory. In fact, Data Science/ML come under Applied Mathematics only. So, to get a deep knowledge, there is no alternative of reading/practicing from book and gaining practical knowledge from the industry.

To start with Data Science, you may follow the sequence

  1. Brush up your Probability, Linear Algebra, Statistics, Optimization knowledge. There are not short cuts in doing this. You have to invest time.
  2. Start with an online article/course. A short book like what I authored will also help
  3. Implement what you learn and gain practical knowledge about the use case
  4. Pick up a good book which covers only the theory part well and deep dive into the theory of what you have implemented
  5. Repeat the same cycle from 2.

Data Science is such an area where reading the hard-core mathematical theory at first may develop some disinterest towards it. It is true that not all the people are very fond of Math/Stat. That’s why starting off with a practical approach by reading some online material or a short book which covers business use cases well always helps.  Specially if you are from normal Software Engineering background, this approach works best.  Later you have to develop a deep understanding of underlying Mathematics. Again, I have to say, there is no short cuts for that.

Q8. Which one is your most used Machine Learning Algorithm in your long span of career?

Linear Regression & CNN

Q9. What kind of challenges have you faced in AI and Data Science Industry?

Biggest challenge is bringing the Data Science awareness. Many organizations start with lot of enthusiasm, but after certain period when it is the time to see the result in real environment, people lose tracks. It happens due to lack of awareness and benchmarking about Data Science. People misunderstands the life cycle of AI/ML based applications. It involves fair amount of research activities and it may happen that after investing sufficient time nothing came out. It does not go like traditional water-fall or even modern-day agile methods. It has its own way. People have to be open about it.  Ultimately, Data Science practice adds lot of values in any org of today’s world. It can bring significant growth in long run.

I have observed, many times Data Scientist/ML practitioners are misunderstood by the people & senior management. It is driven by fear and lack of awareness about the subject which is not desirable at all.

It is high time that Industry should invest time in benchmarking and bringing process in this area. In some places this activity has already started but still lot of things need to be done.

I feel Data Scientist like me should come forward and educate people about the subject through their books, blogs, videos, lectures and many other possible ways. After all it is the future.

Q10. Please tell us about your online blogs?

My blogs are there in medium: https://medium.com/@avisheknag17