Putting Data Science in Production - Mikio Braun
00:00 Tools landscape which support putting ML models in production
00:59 Core of ML workflow & tools
02:50 Meaning of "in production"
05:10 Training more complex models. Tools & why using them
07:10 Tools which support ML workflow
09:00 Tools to scale out in production
12:00 Automation ML workflow
16:25 Real-time Application: How to
19:00 Real-time Application: Dealing with Online & Offline features
20:56 Real-time Application: Architecture
22:45 Tools for the production phase
24:35 Labelling data
26:45 Tools landscape for ML workflow summary
28:00 Challenges & workarounds
32:00 1st question: How to choose the right tools
36:10 2nd question: Popular tools to tackle ML problem
37:00 3rd question: Most important tools?
40:02 4th question: Where would AWS Athena/Presto fit best in the workflow? Tools recommendation?
41:28 5th question: Automation tools recommendations
42:44 6th question: Can one person handle all these tools/tasks ideal
46:48 7th question: Suggested roles in a data team to cover different workflows?
48:50 8th question: Favorite tools & platforms of Mikio
50:27 9th question: Different patterns of models deployment
Every year, the number of tools that promise to take care of everything that's needed to put ML into production is growing, but which tool is really the right tool for the job and what does what? In this talk I'll be taking a look from the inside out, starting with an ML algorithm and gradually adding layers to give a bit of structure to this area. At the end, you probably have a better understanding what kinds of tools exist and what you need when.
Links:
Slides: https://www.slideshare.net/mikiobraun...
Mikio's Twitter: / mikiobraun
LinkedIn: / mikiobraun
Blog: https://margint.blog/
Making figures: https://margint.blog/2018/05/27/how-i...
ML in practice book: https://book.mlinpractice.com
Thanks to Hong-Ngoc Emily Tran ( / emily-ngoc-tran ) for timecodes!
Join DataTalks.Club: https://datatalks.club/slack.html