Tech Industry
7h
1752
Tech companies to avoid as a white guy?
Tech Industry
9h
2992
Be a wh0re
India
2h
288
People who like Modi, are you okay with your country’s leader using words like “infiltrators” for a community?
Tech Industry
Yesterday
1234
Disgusting Indians in Canada
Tech Industry
2h
534
Single people, would you hire an escort to accompany you to company party?
How does one become a machine learning engineer? What skills are needed besides ML? What specifically should one study for software engineering?
For large scale ML systems, I like the deep dives of: www.machinelearningatscale.com
There are two kind of ML roles :- 1. FAANG and other top tier companies like Uber, Snap, pins, Microsoft who do real ML work and need to really innovate for processing huge amount of data. 2. Everyone other company who believes Linear Regression is good enough , using XGBoost makes you a ML engineer. If you are ok to be engineer of 2nd kind, you can do following: 1. Complete Andrew Ng’s ML course on Coursera. This is absolute rock bottom baseline for ML engineers . De facto standard. 2. You can then subscribe to DeepLearning.org specialisation track. The Scepialization covers everything you need to know to become a successful ML engineer- from writing case studies to project reports, to actually tuning models. By doing this, you will start understanding basic ML concepts. You will start to appreciate complexity of model selection, identify success criteria correctly and know when ML is not a right solution. Basically you will have all the experience needed for applying to non-FANG roles. Beyond this, if you want to build a better ML resume, you should enrol in Linear Algebra and Statistics course. You should be able to understand how Bayes probability works. What is multi variate analysis and how different probability distributions work. Once you are comfortable with Calculus of Probability, you can refer to the modern literature of ML. Pick one area of specialisation - recommendations, personalisation, fraud, image and text classification, generative art, whatever. And improve in that one area. As regards to ML specific software engineering , you should definitely read about MLOps. Generally speaking , Phd holders don’t have any idea how to write good code. They would create good models but have no idea how to make it executable under heavy load, GPU tuning, optimising train and test cycle, etc. MLOps brings software engineering to ML world. Another area to explore is Data Engineering - 20% is model building ,80% is data wrangling for any ML project. You can read about feature selection, pipeline building and Exploratory Data analysis. These technically aren’t ML, but a ML project cannot succeed without Data Engineering. Good part is this area needs just High School calculus. If you understand how PCA works, basically you are over qualified.
Wow Amazing reply. Thank you!
Thank you so much for responding! I did my undergrad in applied math and currently doing my ms in data science. I built a model and used docker to put it in a container so I can distribute it to other teams. Unfortunately, I didn’t use any unit testing or software engineering best practices.