Scanning all new published packages on PyPI I know that the quality is often quite bad. I try to filter out the worst ones and list here the ones which might be worth a look, being followed or inspire you in some way.

Vertica-ML-Python simplifies data exploration, data cleaning and machine learning in Vertica.

Easy data preprocessing and data augmentation for deep learning models

High performance distance histogram calculation framework for CPUs and GPUs. CADISHI \– CAlculation of DIStance HIstograms \– is a software package that enables scientists to compute (Euclidean) distance histograms efficiently. Any sets of objects that have 3D Cartesian coordinates may be used as input, for example, atoms in molecular dynamics datasets or galaxies in astrophysical contexts. CADISHI drives the high-performance kernels pydh (CPU) and cudh (GPU, optional) to do the actual histogram computation. The kernels pydh and cudh are part of CADISHI and are written in C++ and CUDA.

This project aims to provide some offline simulators for training and testing recommender systems of education.

Python library of NLP functions originally collated by Equinor Knowledge and AI Data Science team.

Neural Question Answering at Scale. Use modern transformer based models like BERT to find answers in large document collections

An open source library for statistical plotting

Add Recycle bin to your jupyter notebook.

Make Neural Models as APIs for solving more complicated AI problems

The Tanystes Deep Learning Framework

A compact Python toolbox for transfer learning. transfertools is a small Python package containing recent transfer learning algorithms. Transfer learning strives to transfer information from one dataset, the source domain, to a related dataset, the target domain. Several constraints and assumptions can be placed on the domains, inspiring different algorithms to do the information transfer. The package contains four transfer learning algorithms.

Machine learning tools for the Geoscience Australia uncover project