How I broke into Data Science without master's or Ph.D.
Simple, honest and straight, engaging practical guide with a learning path.
Roadblock: It is hard to break into DS without a relevant degree or many years of relevant commercial experience. However, I broke into it. So, you too can :-)
Math Background (Statistics, Calculus, Probability)
You are lucky if you have been very good at Mathematics in your school/college/under grad and have shown intuitive capabilities.
IMO, Mathematical/intuitive understanding is necessary to be a good Data Scientist
Even if you were good at Math, it would still take 1–2 years to refresh, review and understand the proofs required for DS and develop the intuition
…
If you had an average Math history, you could still learn now with mini courses, but it would take more and more time. Rushing will not help
Online Materials:
So many. Thanks for the confusion
I would suggest a top few which I felt are better in quality.
- Penn State Stat course — Brilliant! Thanks for them. Whenever I needed to derive/understand the proof of something, I refer to this.
- Probability Course — Brilliant Again! High Quality! Written by a Ph.D. author. Definitions are thorough and succinct. I prefer the author’s Mathematical notations.
- Business Analytics —Book by IIM professor. Monstrous book and a wide variety of topics including a number of problems. Don’t expect detailed proofs. You need to refer somewhere else for proofs.
- YouTube/Medium/educative.io: You need to supplement your studies with YouTube videos — Stat Quest, Khan academy,Serrano Academy
!!! Alert !!!: More articles more confusions! I can give at least three definitions of ‘percentiles’ in different contexts! Hmm, hang on for my articles.
Why to understand proofs? — To understand and remember the concept correctly and develop the right intuition.
Tip: Don’t be afraid to spend time on deriving something by yourself or understanding the concepts. I remember to have spent so multiple days understanding a single concept!
Post-graduation?
Udacity Nanodegree: Awesome! I would rate it world class! Projects and datasets by the industry’s leading ML engineers/researchers. I did Data Scientist and Deep Learning Nano degrees from Udacity :-)
Masters/Ph.D.: Take your time and money! Industry, as of now, prefers people with degrees. I did not take this path though.
I recently started on a masters on a different but on a related field:
Caution for experienced folks —Your life is going to be tough! The MOOC courses would teach you top class content, present a certificate. There ends the story. There is no guarantee** that the industry would give you a job based on the certificate leaving your experience/past domain behind. Try your luck in a startup company
** Some paid courses offer refund if not placed in certain days after qualification.
Tip: Try to get some ML/DS experience in your current/surrounding projects within your current organization. Nowadays organizations are modernizing their fleet of products with AI/ML in one or the other way. There is every opportunity round the corner, and you should grab it. Else, your degree/certificate would become obsolete soon.
Tip: Maintain a GitHub repo with your portfolio of projects! So, you can upload and showcase your work to potential employers. I was very lazy and conservative before Udacity made me maintain a GitHub repo :-D
For newbies — Life is somewhat easier if you do any related master's course
Python:
Last but not least. The WOW factor when you come from a different language background and enjoy the beauty, simplicity and power of Python and its ecosystem especially for data analysis. Please let me know in comments how you felt switching to Python
Caution: Some people start to think being extremely good at Python makes you a good data scientist. This is the wrong path unless you want to be a developer. Data Science is Science in the first place. Bring the scientific acumen out of you
Success story: I got two offers — (Sports Betting) Data Scientist and Principal ML Engineer (Education software). I chose the Data Science path since I love toying with data and making inferences through new models/establishing statistical relationships (correlation, causation etc.,) between variables from both noisy data and controlled data
Key message: Patience is the key. If you are good at it but someone is not even considering your application, be happy but keep applying.
If you broke into DS/ML eventually, then a final piece of advice :-D
Advice: Many commercial enterprises want output and not research. That is — they are interested in applied AI. If you want to do more research on the subject just out of passion, then you should do a Ph.D./ Masters by Research/Quant job.
