0
$\begingroup$

MLOPs has been gaining traction and a lot of fortune 500 companies are creating team specifically for MLOPs.Can anyone help me understand?

  1. Why MLOPs is gaining so much traction?
  2. How it is different from DevOps?
  3. What are the tools used for MLOPs?
  4. How can i create a strategy for MLOPs?
  5. How to get started?
$\endgroup$
1
  • $\begingroup$ While this is a too broad question for a Stack Overflow site, I can point you toward "Hidden Technical Debt in Machine Learning Systems" which was written by Google in 2015 and published in NeurIPS. MLOps can broadly be seen as a set of practices created to counter the issues mentioned in that paper. $\endgroup$
    – Avatrin
    Feb 11, 2022 at 21:13

1 Answer 1

2
$\begingroup$

Why MLOPs is gaining so much traction?

Because MLOps is about management of the lifecycle of machine learning, and without proper management any machine learning project would just fail in real production.

How it is different from DevOps?

The difference lies in the difference between machine learning and traditional programming: 1) machine learning projects are uncertain; 2) machine learning projects are iterative(iterative in each step and iterative in the whole lifecycle); 3) machine learning projects are data-hungry.

What are the tools used for MLOPs?

  1. Programming language: Python, R, C/C++
  2. Data processing toolset: Jupyter Notebooks(Colab or Kaggle kernel), Pandas, Numpy, Sci-kit learn, Matplotlib, NLTK, SpaCy
  3. Modeling toolset: TensorFlow, PyTorch, Keras
  4. API / interface: AutoML
  5. Platform: MLFlow, AirFlow, Kubeflow

How to get started?

Read books or watch lectures.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.