Thanks for this reading. It's amazing focus to build ML products. The conversations with the stakeholders is essential (totally agree), sometimes takes more time to catch certain details in the business problem.
I wanna ask you two questions:
1. Do you recommend use virtual environment in python with poetry as well? (I know today this tool to manage the requirements). The majority of the time you can create a Jupyter notebook in Anaconda without virtual env and that's all.
2. What's the final product to our customer? (Could be the Jupyter notebook or the deploy)
1) Poetry is a tool that helps you both create a virtual env, where you develop your code, AND to package it, so you can easily share it with others. I am no big fan of Anaconda, as it comes with a lot of predefined packages that I hardly use.
2) The final product has to bring business value, and for that, you have to at least deploy it at small scale, so you can test it end-2-end. Jupyter notebooks are enough when the client only wants to assess if investing more in ML makes sense or not. However, once your notebook produces something meaningful, they immediately ask you to "somehow" push it to production.
Going from notebooks to ML products is one of the gaps I see in the market. It is also a great opportunity for ML engineers to stand out from the crowd.
This is precisely what you can learn, step-by-step, in the Real-World ML Tutorial.
Hello Pau,
Thanks for this reading. It's amazing focus to build ML products. The conversations with the stakeholders is essential (totally agree), sometimes takes more time to catch certain details in the business problem.
I wanna ask you two questions:
1. Do you recommend use virtual environment in python with poetry as well? (I know today this tool to manage the requirements). The majority of the time you can create a Jupyter notebook in Anaconda without virtual env and that's all.
2. What's the final product to our customer? (Could be the Jupyter notebook or the deploy)
Thanks a lot.
Best regards,
Michael
Hi Michael,
Thanks for your feedback.
Let me answer your questions:
1) Poetry is a tool that helps you both create a virtual env, where you develop your code, AND to package it, so you can easily share it with others. I am no big fan of Anaconda, as it comes with a lot of predefined packages that I hardly use.
2) The final product has to bring business value, and for that, you have to at least deploy it at small scale, so you can test it end-2-end. Jupyter notebooks are enough when the client only wants to assess if investing more in ML makes sense or not. However, once your notebook produces something meaningful, they immediately ask you to "somehow" push it to production.
Going from notebooks to ML products is one of the gaps I see in the market. It is also a great opportunity for ML engineers to stand out from the crowd.
This is precisely what you can learn, step-by-step, in the Real-World ML Tutorial.
https://realworldmachinelearning.carrd.co/
Hope this helps,
Pau