Collections/The ML Stack That Actually Ships
Machine Learning · collection

The ML Stack That Actually Ships

The frameworks, serving engines, and libraries that real machine learning work gets built on — from training the model to putting it in front of users.

S
Created by Sanjeet Pal Singh
updated 4 hours ago
0
followers
5
projects
0
likes
0

Projects in this collection

Ranked by the curator, with their reason for including each one.

1
pytorchPython
New102.8K GitHub starsOther
Why Sanjeet Pal Singh included it: The default research framework for modern deep learning, where its eager, Pythonic autograd made experimentation feel like ordinary programming and most new model code lands here first.
2
transformersPython
New164.9K GitHub starsApache-2.0
Why Sanjeet Pal Singh included it: The common vocabulary for state-of-the-art models across text, vision, and audio — pretrained architectures you can load and fine-tune without reimplementing a paper.
3
scikit-learnPython
New67.2K GitHub starsBSD-3-Clause
Why Sanjeet Pal Singh included it: The workhorse for everything that isn't a neural network, with a consistent fit/predict API and sane defaults that still solve a huge share of real-world prediction problems.
4
vllmPython
New91.1K GitHub starsApache-2.0
Why Sanjeet Pal Singh included it: Closes the loop from trained model to production by making large language model inference fast and memory-efficient enough to serve at scale.
5
tensorflowC++
New198.9K GitHub starsApache-2.0
Why Sanjeet Pal Singh included it: The end-to-end platform that carried ML into production first, with a deployment ecosystem spanning servers, mobile, and the browser that still anchors a great deal of shipped ML.