Every day there is a new framework, a new implementation, or a new tool that makes life easier. It is indeed very hard to stay updated, and even doing it would mean giving up a lot of time we could invest in doing something else.

Of course, we are not meant to be always updated with the newest discoveries, or the new minor release of a specific library but sometimes it is necessary, interesting or simply we are curious to find something new!

Let’s start!

Texthero: Text preprocessing, representation, and visualization from zero to hero. Apply tf-idf , tokenize, do PCA in a pipeline-oriented way.

Google Data Studio: your to-go frontend. Create dashboards, reports, and analyses in a Google Docs way. Just plug your DB, upload your CSVs, and get started for free.
Deepnote: jupyter notebooks on steroids. Collaborate, code reviews, better plotting, support for AWS S3, MongoDB, and many more. All in your browser.
Streamlit: The fastest way to build data apps. An alternative to Google Data Studio. Create python-based web apps, visualizations, and reports.

Coming from R and switching to python? Try plotnine: an implementation of a grammar of graphics in Python, based on ggplot2.
pivottablejs: drag ‘n drop pivot tables in Jupyter Notebooks.
RISE: turn your notebooks into a reveal.js-based slideshow.
gmaps: Google Maps-based visualization library: create beautiful and interactive maps and heatmaps.
flair: a very simple framework for state-of-the-art NLP. Backed by Zalando in Berlin.
light fm: a python implementation of popular recommendation algorithms.
ds-cheatsheets: a huge collection of cheatsheets, from python to R, including SQL.
Scraper.AI: a web scraper that actually works.
AlwaysAI: deploy computer vision model to edge devices such as Nvidia Jetson, Raspberry PI in minutes. Their catalogue covers different pre-trained models, from object segmentation to pose estimation.
Notion: The unopinionated note-taking app. Use Markdown, create tables, list, canvas, and even kanban boards.

Ah, they also provide a Python API:

Weights & Biases: while training deep learning models, often could happen that results from experiments get lost, overwritten or it is difficult to keep track. Weights &Biases help you keep track of model training, experiments, just by adding a few lines of code.

Machine learning without code? Obviously AI: probably the next step in AutoML, it is enough to upload (or connect) your data, pick your target, and ObviouslyAI will do the rest making the ML process accessible to anyone. They also generate a decision tree for you, helping to provide an explainable model.

ML Playground: play with different algorithms, add neurons, remove layers, draw your data, or upload yours!
Papers with code: exactly what it says, find papers with attached together the GitHub repo. Ready to be forked!
Clever Grid: Get a 1-core GPU along with 250GB of training data for about 10€ per day.
AWS DeepRacer: train your self-driving (model) car, compete with other people on famous F 1 tracks such as Circuit de Barcelona-Catalunya. You can also buy a hardware version of the DeepRacer car on Amazon.

Music Time for Spotify: a VSCode editor plugin that discovers your most productive music to listen to while you code.
gspread_dataframe: ever needed to send a pandas dataframe to Google Sheets?
Kite: when AI meets code autocompletion and suggestions. They provide plugins for mayor Python IDE such as VSCode, Pycharm, and Spyder.

PuLP: ever interested in the word of integer programming and linear optimization? interested in problems such as production optimization or multi-armed bandit? Check their case studies.
Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems the book every data scientist should have. It covers from the basics to advanced topics, extremely useful and written with the idea of a manual, hence a book you would return in the future to brush up. Plus even my dog got a copy.

datatau: Hacker News for data science.
Deta: a cloud provider with a particularly generous free tier!
Looking for a side project? Find side projects you are interested in and take part in! Solodoers.
cookiecutter-data-science: a project bootstrapper for data science. Because data science code quality is about correctness and reproducibility.
tqdm: because we always wanted to have a progress bar in for loops.
ELI5: visualize and debug various Machine Learning models, from black-boxes to explainable AI.

gpxpy: You know you can export your favorite’s running app data into a .gpx file? Those files can be parsed into pandas (maybe a data science project for your portfolio?) I once did something similar exporting data from a sailboat trip:

Getting Your First Data Science Job: A Free 70-page book on data science careers with expert advice
GluonTS: probabilistic time series modeling in the Amazon way. Based on mxnet.

Lifelines: a Python library that implements common survival analysis models. Survival analysis is widely used in predicting things such as how likely an event occurs at a specific time, for example, that a customer will unsubscribe to our service.
tensor-house: A collection of reference machine learning and optimization models for enterprise operations, really interesting to learn how different machine learning models can be used together to solve different real-life problems.
Gradio: create easy-to-use UIs for your models, very interesting for showcasing models predictions, from NLP to images and regressions.