Read on Twitter
One day one of my mentees asked me: "How to break in this Machine Learning / Data Science world?" She was a Software Engineer in a Big Tech who wanted to do a career shift.
So, I'll try to do my best here and summarize our discussion
It depends from the business or research area. Data Science /ML is like Software Engineering and spans across industries, roles, fields. You have to think where you want personally to be more involved in order to excel to this specific area that you want.
NLP/text mining, IR/search engines, recommendations, personalization, statistical analysis of structured data like A/B tests, classifying or clustering or predicting something
Do you consider yourself a builder of something? Someone who empowers others through your technical infrastructure or your insights? A story-teller? A strategist?
Do you take energy when you are deep in your code? When you're building something or debugging or make sense of data? Or when you're discussing valuable insights with others? Or when you define strategies?
But more importantly: Will you be ok with the fact that the job title is not going to define *you*? Are you going to be ok with the fact that Data Science requires a bit of all of the above?
Check some illustrative role titles such as Machine Learning Researcher /Engineer, Data Engineer, Data Scientist, BI Analyst, Product Manager (we did this exercise together back then) to see where do you believe will fit mostly
Then, check trainings & certifications in order to apply in action the things that you study (a great way to learn something -> action / involvement in projects / certifications)
You could participate in a Kaggle competition to gain a 360 view of stages: Business understanding (somewhat) - data understanding - data preprocessing (this is a heavy one 70% sometimes of your time!) - EDA - feature engineering /extraction or selection - build model - test
Do stuff with R or Python -
Start with EDA. This is the first skill - to understand data even with summary statistics / visualizations. Then get a bit deeper on algorithms.
Start with EDA. This is the first skill - to understand data even with summary statistics / visualizations. Then get a bit deeper on algorithms.
Basic areas: Supervised learning (classification algorithms,) unsupervised learning (clustering algorithms), regression, feature selection (eg dimensionality reduction).
Take courses linked with the role you like: Marketing & customer analytics -> Statistics, Data Mining. Recommendation systems -> distributed systems (map-reduce), algorithms & complexity, data structures (to deal with scalability of your infrastructure), IR, tf-idf concepts
My mentee wanted to do the shift and stay in Industry (not Academia / Research)
Then, we got into the next round of discussions: Business acumen and influence
It's a tough dance between: What the business needs, what motivates you, and... culture.
It is never enough to emphasize the importance of communication and presentation skills. The DS is part of a interdisciplinary team. You have to directly or indirectly influence with your insights executives, clients, and business people with not technical background
I love the motto: "Correlation does not imply causation". But I love more the fact that I made it the fav quote of some senior executives with only traditional business background behind them.
